Fatal Accidents among Elite Mountaineers: A Historical Perspective from the European Alps

Abstract

Weinbruch, Stephan, and Karl-Christian Nordby. Fatal accidents among elite mountaineers: a historical perspective from the European Alps. High Alt. Med. Biol. 11:147–151, 2010.— The lifetime risk of a fatal mountain accident among elite European alpine mountaineers and its time trends are determined by studying a fixed cohort of 390 elite mountaineers listed in the Encyclopaedia of the Alps (Hiebler, 1977). At publication of the encyclopaedia, 158 individuals were still living and were followed up until the end of 2008. The crude lifetime risk of a fatal accident for elite mountaineers is 0.203 [95% confidence interval (CI), 0.165 to 0.246). The difference in mortality between male (0.207; 95% CI: 0.168 to 0.251) and female mountaineers (0.118; 95% CI: 0.033 to 0.343) is not statistically significant. No fatal accidents occurred among elite mountaineers born before 1820. For the birth cohort from 1820 to 1949, the lifetime risk of a fatal accident (male mountaineers only) increased with time from 0.069 (95% CI, 0.019 to 0.220) to 0.375 (95% CI, 0.212 to 0.573). For all time strata, the highest risk of a fatal mountain accident was observed at an age of 30 to 39 yr. The high mortality among elite mountaineers clearly demonstrates that the limits of human performance are reached by these activities. The high risks should be communicated and should motivate risk-reduction efforts for this highly exposed subgroup of mountaineers.

Introduction

Although mountaineering (rock climbing, mixed and ice climbing, high altitude climbing, ski mountaineering) is generally perceived as a high-risk activity, quantitative risk estimates are still sparse. There are basically two different kinds of studies: (1) reports of alpine clubs or mountain rescue organizations, which are usually not published in the scientific literature, and (2) scientific papers that study risks at locations with reliable records of mountaineering activity, for example, national parks or high altitude climbing requiring permits (Weingart, 1980; Pollard and Clarke, 1988; Schussman et al., 1990; Lattimore, 1993; Malcolm, 2001; Firth et al., 2008; McIntosh et al., 2008). In the first case, average values for large inhomogeneous populations (with respect to exposure) are determined for different activities, including climbing, mountaineering, hiking, and ski mountaineering. Subgroups at higher risk are usually not identified in these reports. In the second case, the investigations are often restricted to small areas (a national park or a single mountain), and it is difficult to generalize the findings of these studies. There is only one prospective cohort study on morbidity and mortality in a population of mountain climbers (Monasterio, 2005). At 4-yr follow-up, four fatal mountain accidents were observed among the 46 participants. Recently, a study of morbidity in mountaineers (British mountain guides) was published by Hillebrandt (2007). Because of the small sample size, mortality was not included in this investigation.

Accident risks are certainly not distributed homogeneously among mountaineers. Activities with the highest risk include climbing (rock, ice, and mixed terrain) at a high-difficulty level (relative to the achievements of a given time), high altitude mountaineering, and expeditions to polar regions. Quantifying the lifetime risk of fatal accident for the most exposed mountaineers has not been attempted so far. A major difficulty encountered in this task is the lack of detailed records on climbing activity. Such data are only available for high altitude climbing in the Himalayas (Salisbury and Hawley, 2007). Therefore, the selection of a sample of elite mountaineers cannot be based on detailed exposure data (e.g., time at risk). Instead, inclusion in the present study was based on the general recognition of the achievements of a person, which of course depends on subjective judgment. In the present contribution, the lifetime risk (incidence proportion) of a fatal mountain accident is estimated for a sample of elite mountaineers from the European Alps. Special emphasis is placed on the time trends of this outcome. Because of the uncertainties introduced by the selection procedure, the obtained lifetime risk of fatal mountain accident is regarded as an order of magnitude estimate.

Methods

A fixed cohort of 390 elite mountaineers was studied. This cohort contains all mountaineers listed in the Encyclopaedia of the Alps (Hiebeler, 1977). Selection for inclusion into this encyclopaedia is based on outstanding achievements in climbing in the European Alps (first ascents, first winter ascents, frequent climbs of high difficulty). Thirty-one persons listed for other reasons than high-performance mountaineering (e.g., development of mountaineering equipment, achievements in mountain rescue) were not included. At publication of the encyclopaedia, 158 of the mountaineers listed were still living and were followed up until the end of the year 2008. The outcome investigated was fatal mountain accident from all causes. Confidence limits for the risk data were calculated using the Wilson approximation for a binomial distribution (Wilson, 1927), as recommended by Rothman (2002). The homogeneity of samples was tested with χ-square statistics (using STATGRAPHICS^® version 5, Warrenton, VA, USA.).

Results

The characteristics of the investigated cohort (n = 390) are summarized in Table 1. From the 158 mountaineers still living at the year of publication of the encyclopaedia, 5 persons (4 male, 1 female) were lost during follow-up.

Table 1.

Characteristics of the Cohort Studied (n = 390)

Gender
Male	Female
372	18
Year of birth	n^a	Country	n^a
<1800	8	Austria	79
1800–1819	6	France	26
1820–1839	29	Germany	80
1840–1859	36	United Kingdom	22
1860–1879	46	Italy^b	86
1880–1899	47	Switzerland	80
1900–1919	95	Other^c	17
1920–1939	95
≥1940	28

Number of mountaineers.

Including South Tyrol.

Countries with less than 10 persons (Yugoslavia 9, United States 4,

Belgium 2, Czechoslovakia 1, Spain 1).

From the remaining 385 elite mountaineers, 78 experienced a fatal mountain accident (incidence proportion 0.203; 95% CI, 0.165 to 0.246). The incidence proportion for male mountaineers (0.207; 95% CI, 0.168 to 0.251; n = 368) is higher compared to female mountaineers (0.118; 95% CI, 0.033 to 0.343; n = 17), although not reaching statistical significance (p = 0.373). Because the number of women in the cohort is rather small, the stratified analysis is restricted to male mountaineers.

The time dependency of the lifetime risk of a fatal mountain accident is shown in Table 2. For mountaineers born before 1820, no fatal mountain accident is observed. For mountaineers born after 1820, there seems to be a steady increase in the risk with time (except for the birth cohort 1840 to 1859).

Table 2.

Lifetime Risk of a Fatal Mountain Accident for Male Mountaineers Stratified by Year of Birth

Year of birth	N^*	Fatal accident	Incidence proportion	95% CI
<1800	8	0	0.000	0.000–0.324
1800–1819	6	0	0.000	0.000–0.390
1820–1839	29	2	0.069	0.019–0.220
1840–1859	36	8	0.222	0.117–0.381
1860–1879	44	7	0.159	0.079–0.294
1880–1899	46	7	0.152	0.076–0.282
1900–1919	90	20	0.222	0.149–0.318
1920–1939	85	23	0.271	0.188–0.373
1940–1949	24	9	0.375	0.212–0.573

Number of mountaineers.

CI = confidence interval.

The lifetime risk of a fatal mountain accident appears to be lowest for mountaineers from Italy and Switzerland and highest for British mountaineers (Table 3). However, the differences among the different nationalities are not statistically significant (p = 0.787).

Table 3.

Lifetime Risk of a Fatal Mountain Accident for Male Mountaineers Stratified for Nationality

Country ^a	n^b	Fatal accident	Incidence proportion	95% CI
Austria	76	17	0.224	0.145–0.329
France	22	5	0.227	0.101–0.434
Germany	77	17	0.221	0.143–0.325
United Kingdom	20	6	0.300	0.145–0.519
Italy^c	84	15	0.179	0.111–0.274
Switzerland	77	13	0.169	0.101–0.268

Countries with ≥20 mountaineers included in the cohort.

Number of mountaineers.

Including South Tyrol.

The age distribution of fatal mountain accidents and the corresponding risks for the different age groups are shown in Table 4 stratified by three time intervals. Mountaineers born before 1820 are excluded; no fatal mountain accident was observed for this group. For the risk estimates shown in Table 4, competing risks are taken into account. As mentioned previously, the risk of a fatal mountain accident increases with calendar time. This increase of the risk with time is clearly visible for the two age intervals 20 to 29 and 30 to 39 yr. For the other age intervals, no clear time trend of the risk estimate is observed. The risk estimate for the age group 60 to 69 yr of the birth cohort 1920 to 1949 is uncertain, because 18 persons of this age were still living at the end of the follow-up period. No fatal mountain accident was observed for an age ≥70 yr. In all three time strata, the highest risk is observed for the age interval of 30 to 39 yr.

Table 4.

Life-Table for Death from Mountain Accidents (Male Mountaineers)

Age at accident (yr)	Fatal mountain accident	Deaths from competing risks	Survivors at end of interval	Risk	95 % CI
Year of birth, 1820–1889; 135 persons
0–19	0	0	135	0.000	0.000–0.028
20–29	4	1	130	0.030	0.012–0.074
30–39	7	3	120	0.054	0.026–0.107
40–49	5	3	112	0.042	0.018–0.094
50–59	4	9	99	0.036	0.014–0.088
60–69	2	14	83	0.020	0.006–0.071
Year of birth, 1890–1919; 110 persons
0–19	0	0	110	0.000	0.000–0.034
20–29	7	3	100	0.064	0.031–0.126
30–39	9	5	86	0.090	0.048–0.162
40–49	3	2	81	0.035	0.012–0.098
50–59	1	3	77	0.012	0.002–0.067
60–69	2	11	64	0.026	0.007–0.090
Year of birth, 1920–1949; 109 persons
0–19	0	0	109	0.000	0.000–0.034
20–29	8	0	101	0.073	0.038–0.138
30–39	18	2	81	0.178	0.116–0.264
40–49	3	1	77	0.037	0.013–0.103
50–59	2	4	71	0.026	0.007–0.090
60–69	1	(4)^a	—	(0.014)^a	(0.002–0.076)^a

Uncertain; 18 persons of an age between 60 and 69 yr were still living at the end of 2008.

About two-thirds of the fatalities (53 out of 78) occurred during an activity that can be designated as high-level mountaineering (at the year of incidence) restricted to elite mountaineers. The distribution of the fatal accidents among the different disciplines of mountaineering is shown in Table 5. Because most persons listed in the Encyclopaedia of the Alps were engaged in several of the disciplines shown in Table 5 and detailed records of their activities are not available, no risk estimate can be given for the different disciplines of mountaineering.

Table 5.

Distribution of Fatal Accidents among Different Disciplines (Male Mountaineers)

Discipline	Fatal accident
Rock climbing	19
Mixed and ice climbing	31
High altitude mountaineering	12
Ski mountaineering	13
Hiking	1

n = 76.

Discussion

Validity

Selection bias

Inclusion criteria for the Encyclopaedia of the Alps are outstanding achievements in climbing in the European Alps: first ascents, first winter ascents, and frequent climbs of high difficulty (relative to the standards of a given time). Because these criteria depend on subjective judgment, it may be suspected that the risk estimates are strongly influenced by selection bias. The most obvious possible source of selection bias is preferential inclusion in the encyclopaedia of people who were killed in a mountain accident in order to keep remembrance alive. However, the magnitude of this remembrance effect can be quantified, because a substantial fraction of the mountaineers was still living at the 1977 publication of the encyclopaedia. For this group, the remembrance effect does not, of course, exist. A first estimate of selection bias owing to the remembrance effect is obtained by comparison of the observed versus expected number of cases of fatal mountain accident occurring after January 1977 (the last record of a fatal mountain accident included in the encyclopaedia is from January 17, 1977). The expected number of cases is obtained from a survival analysis and using the age distribution of the mountaineers still living at publication (i.e., the age distribution in 1977). In these calculations it is assumed that the age distribution of fatal mountain accidents (Table 4) of the birth cohort 1920 to 1949 can be applied. For the 141 elite mountaineers (only male) still living in 1977, the expected number of fatal accidents until the end of 2008 is 8.2; the observed number of 7 fatal accidents is close to this value. If the observed difference is completely attributed to the preferential inclusion of people being killed in a mountain accident, an upper limit for the remembrance effect is obtained. In this case, the lifetime risk would be overestimated by approximately 18% (relative).

A second possible source of selection bias is the well-known effect of survival. To be included into the present cohort requires that an individual survived several years with significant achievements in mountaineering. The exact duration of this period of immortal person-time is not known, but it is likely to be at least 5 to 10 yr. When a risk estimate is determined for elite mountaineers who are individuals with the highest exposure level and long exposure time, the survival effect is inherent per the definition in the present study. Consequently, the estimated lifetime risk of fatal accident for elite mountaineers cannot be used to assess the mortality rate (cases per unit time of activity) for climbing at a given level of exposure, expressed for example as degree of difficulty. For this purpose, detailed records of climbing activity indicating time at risk would be necessary. Unfortunately, such data are not available.

Loss to follow-up, which is often a significant source of selection bias in cohort studies, can be neglected in our case, because this loss was small (5 out of 158 persons).

The influence of competing risks (censoring) was taken into account in the calculation of the age-stratified risk estimates (Table 4). However, the number of elite mountaineers who died of other reasons before reaching the age of 70 is small, and after this age the observed risk of fatal mountain accident is practically zero.

The estimate of a lifetime risk is also not affected significantly by the fact that 57 male mountaineers were still living at the end of the follow-up period, because at this time the cohort members were all, except one, over 60 yr old and have, thus, a low probability of experiencing a fatal mountain accident in the future. Assuming that the age distribution of fatal accidents (Table 4, birth cohort 1920 to 1949) did not change, the number of fatal mountain accidents that is expected to occur after 2008 is calculated to be 0.2, which is close to zero.

Information bias

To reduce misclassification, all cases of fatal mountain accident reported in the original data source (Hiebeler, 1977) were verified using alternative information sources. To accomplish this, we accessed popular mountain books, annual reports of mountain clubs and rescue organizations, and information on the Internet. Information on fatal mountain accidents occurring during follow-up after January 1977 was obtained likewise, as well as by contacting relatives and friends of the victims. In all cases, the event of a fatal accident was judged to be reported accurately. The number of fatal accidents overlooked in the sample is also expected to be small, because information on the cause of death for people not killed in a mountain accident or confirmation that a person is still living (end of 2008) was obtained in almost all cases from more than one source. The exact date of birth and date of death listed in Hiebeler (1977) was found to be wrong in a few cases. These errors were corrected prior to data analysis.

Risk predictors and confounding

The main exposure analyzed is inclusion in the Encyclopaedia of the Alps, which is a proxy for high-level mountaineering at a given time. The risk of fatal accidents is influenced by several other factors, including gender, year of birth, nationality, and age. Reliable information on the socioeconomic status and education was not available for the present study. Unfortunately, confounding was not considered in most previous cohort studies, and only crude risk estimates were given (Weingart, 1980; Pollard and Clarke, 1988; Schussman et al., 1990; Malcolm, 2001; Firth et al., 2008). Gender, age, and nationality of mountaineers were included only in the database for the Nepal Himalaya (Salisbury and Hawley, 2007). In the following, the role of confounding will be discussed briefly.

The point estimate for the incidence proportion of fatal mountain accident among elite mountaineers appears to be much lower for women than for men. According to Robertson (2007), women are usually at less risk for most types of injury. Indeed, for high altitude mountaineering in the Himalayas, a significantly higher mortality was reported for men than for women when all peaks above 6000 m were considered (Salisbury and Hawley, 2007). However, the situation is more complex when the data are stratified into different altitude intervals (in meters). Men were found to have a significantly higher mortality than women only for the 7000ers; for 8000ers and 6000ers, the differences in mortality were statistically insignificant (Salisbury and Hawley, 2007).

The risk of a fatal accident for elite mountaineers strongly depends on the year of birth (Tables 2 and 4). The strong increase of the incidence proportion from the early 19th century to the first half of the 20th century is parallel to the increase in technical difficulty reached by the leading mountaineers of a given time. This parallel development does not necessarily imply causation. An alternative explanation is that the frequency of climbing changes according to birth cohort and age, which depend on factors such as more leisure time and increased economical assets. In addition, ideological factors (e.g., competition between different nations) may have led to more attempts at prestigious and dangerous climbs, for example, the North Faces of Matterhorn, Eiger, and Grand Jorasses in the 1930s or the 8000ers before and after World War II.

In the present study, the incidence proportion of fatal accident observed for elite mountaineers (Table 3) is distributed homogeneously among the different nationalities. In contrast, pronounced differences of mortality between different nationalities were reported for high altitude mountaineering in the Himalayas (Salisbury and Hawley, 2007). However, if only the six countries studied in the present paper are compared, the mortality of high altitude mountaineering (Salisbury and Hawley, 2007) is also distributed homogeneously among the different nations (p = 0.715). Thus, it can be concluded that different national risk cultures do not exist among the individuals investigated in this study, regarding the six specified countries. It should be emphasized here that our study focuses on European mountaineers and on a time period during which Europe was the center of alpinism. In more recent years, however, high-performance mountaineering has emerged in many countries worldwide, and it is not known if the results of the present study apply to these other countries.

The risk of a fatal mountain accident shows a distinct age distribution (Table 4). Despite a general increase of the incidence proportion with calendar time, the age distribution did not change substantially with time. For all three time strata, the highest risk is observed for the age interval of 30 to 39 yr. The lower values for younger climbers are most likely caused by the effect of survival (immortal person-time) discussed previously. This interpretation is supported by the fact that, for high altitude mountaineering in the Nepal Himalaya (Salisbury and Hawley, 2007), the highest mortality is observed for the age interval from 15 to 24 yr. The decreasing risk at an age ≥40 yr most likely reflects a decreasing frequency of high-performance climbs undertaken at this age.

Comparison with literature data

Categorization of exposure, at least into broad classes (e.g., rock climbing, mixed and ice climbing, and high altitude mountaineering), is certainly desirable in order to check whether the risk is distributed homogeneously among the different disciplines of high-performance mountaineering. However, most members of the cohort studied were engaged in several disciplines. Detailed records of their climbing activities are not available for the majority of them, so the risks associated with the different disciplines of mountaineering cannot be quantified.

Risk estimates for mountaineering published in the scientific literature (reports of alpine clubs or mountain rescue organizations are not included) are summarized in Table 6. The fatality risk strongly increases with the altitude of the peaks attempted. The highest values are on the order of a few percent for a single expedition and are observed for high altitude mountaineering in the Himalayas, Karakoram, and Hindu Kush (Weingart, 1980; Pollard and Clarke, 1988; Salisbury and Hawley, 2007; Firth et al., 2008).

Table 6.

Summary of Risk Estimates Published in the Literature

Location	Time period	Mortality (incidence proportion)	Reference ^a
Denali, USA	1978–1987	4.3/1000 mountaineers	1
Denali, USA	1903–2006	3.1/1000 summit attempts	2
Grand Teton National Park, USA	1981–1986	0.6/1000 mountaineers	3
Mt. Cook National Park, New Zealand	1981–1998	0.6/1000 hut nights	4
		1.9/1000 climbing days
New Zealand	2000	4/46 mountaineers at 4-yr follow-up^b	5
Himalaya, Karakoram, Hindu Kush	1946–1978	30.6/1000 mountaineers	6
Summits >7000 m	1968–1987	43.2/1000 British mountaineers	7
Himalaya, Nepal:			8
Summits (m): 6000–6999	1950–1989	10.1/1000^c
7000–7999	1950–1989	29.4/1000
8000–8850	1950–1989	26.9/1000
6000–6999	1990–2006	4.3/1000
7000–7999	1990–2006	11.2/1000
8000–8850	1990–2006	13.2/1000
Mt. Everest, Nepal	1921–2006	16/1000	9

References: (1) Lattimore, 1993; (2) McIntosh et al., 2008; (3) Schussman et. al., 1990; (4) Malcolm, 2001; (5) Monasterio, 2005; (6) Weingart, 1980; (7) Pollard and Clarke, 1988; (8) Salisbury and Hawley, 2007; (9) Firth et al., 2008.

Prospective cohort study.

Above base camp; only members of the expedition (hired personnel excluded).

High altitude mountaineering in these regions before the year 1990, that is, before the era of commercial expeditions, and at peaks higher than 7000 m was limited to elite mountaineers. Therefore, the risk estimates for these undertakings can be compared to the lifetime risk of fatal accident obtained in the present study. The lifetime risk of a fatal accident for elite mountaineers is about 1 order of magnitude higher than the risk connected with climbing a single mountain above 7000 m. The other risk estimates included in Table 6 refer to mountaineering activities that are certainly not at the elite level and are 1 to 2 orders of magnitude lower. The reasons for the remarkably high mortality observed among a sample of mountaineers in New Zealand (Monasterio, 2005) are not known.

Conclusion

The high mortality among elite mountaineers clearly demonstrates that the limits of human performance are reached by these activities. We have demonstrated that the fatality risk increases from earlier to later birth cohorts and that the risk peaks at age 30 to 39 yr in each birth cohort. For the most recent birth cohort studied in the present paper, between 1940 and 1949, the risk estimates cannot be applied directly to contemporary elite mountaineers. However, it is very likely that they are also at much higher risk than persons not active at the elite level. Therefore, the high risks observed should be communicated and should motivate risk-reduction efforts for this highly exposed subgroup of mountaineers.

Footnotes

Acknowledgments

We would like to thank Nico Mailänder (Munich), Josef Kaiser (Vienna), and Hannsjörg Hager (Bolzano) for help in obtaining information on the elite mountaineers followed up. Reviews by D. Hillebrandt and an anonymous reviewer helped to improve the manuscript and are gratefully acknowledged.

Disclosures

The authors have no conflicts of interest or financial ties to disclose.

References

Firth

P.G.

, Zheng

, Windsor

J.S.

, Sutherland

A.I.

, Imray

C.H.

, Moore

G.W.K.

, Semple

J.L.

, Roach

R.C.

, Salisbury

R.A.

2008. Mortality on Mount Everest, 1921–2006: descriptive study. BMJ, 337:a2654.

Hiebeler

1977. Lexikon der Alpen. Bertelsmann-Lexikon-Verlag: Gütersloh, Germany.

Hillebrandt

2007. Occupational health problems of British mountain guides operating internationally. J. Br. Travel Health Assoc., 10:43–47.

Lattimore

1993. Mountaineering emergencies on Denali. J. Wilderness Med., 4:358–362.

Malcolm

2001. Mountaineering fatalities in Mt Cook National Park. NZ Med. J., 114:78–80.

McIntosh

S.E.

, Campbell

A.D.

, Dow

, Grissom

C.K.

2008. Mountaineering fatalities on Denali. High Alt. Med. Biol., 9:89–95.

Monasterio

2005. Accident and fatality characteristics in a population of mountain climbers in New Zealand. NZ Med. J., 118,1208:U1249.

Pollard

, Clarke

1988. Death during mountaineering at extreme altitude. Lancet., 1:1277.

Robertson

L.S.

2007. Injury Epidemiology. >Oxford University Press: Oxford, England.

10.

Rothman

K.J.

2002. Epidemiology. An Introduction. Oxford University Press: Oxford, England.

11.

Salisbury

, Hawley

2007. The Himalaya by the numbers: a statistical analysis of mountaineering in the Nepal Himalaya. <www.himalayandatabase.com>

12.

Schussman

L.C.

, Lutz

L.J.

, Shaw

R.R.

, Bohnn

C.R.

1990. The epidemiology of mountaineering and rock climbing accidents. J. Wilderness Med., 1:235–248.

13.

Weingart

H.R.

1980. Erkrankungen und Unfälle bei Hochgebirgsexpeditionen. Dissertation. Ludwig Maximilians Universität: Munich, Germany.

14.

Wilson

E.B.

1927. Probable inference, the law of succession, and statistical inference. J. Amer. Stat. Assoc., 22:209–212.