Content uploaded by Darrell Bonetti
Author content
All content in this area was uploaded by Darrell Bonetti on May 08, 2017
Content may be subject to copyright.
Sea-Level Exercise Performance
Following Adaptation to Hypoxia
A Meta-Analysis
Darrell L. Bonetti and Will G. Hopkins
Institute of Sport and Recreation Research, AUT University, Auckland, New Zealand
Contents
Abstract................................................................................. 107
1. Methodology .......................................... ............................... 109
1.1 Study Selection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
1.2 Data Extraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
1.3 Meta-Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
2. Results ............................................................................... 116
2.1 Exercise Performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
2.2 Physiological Measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
3. Discussion............................................................................. 121
4. Conclusions........................................................................... 125
Abstract Adaptation to living or training in hypoxic environments (altitude train-
ing) continues to gain interest from sport scientists and endurance athletes.
Here we present the first meta-analytic review of the effects on performance
and related physiological measures following adaptation to six protocols of
natural or artificial hypoxia: live-high train-high (LHTH), live-high train-low
(LHTL), artificial LHTL with daily exposure to long (8–18 hours) con-
tinuous, brief (1.5–5 hours) continuous or brief (<1.5 hours) intermittent
periods of hypoxia, and artificial live-low train-high (LLTH).
The 51 qualifying studies provided 11–33 estimates for effects on power
output with each protocol and up to 20 estimates for effects on maximal
oxygen uptake ( .
VO
2max
) and other potential mediators. The meta-analytic
random-effect models included covariates to adjust for and estimate moder-
ating effects of study characteristics such as altitude level and days of
exposure. Poor reporting of inferential statistics limited the weighting factor
in the models to sample size. Probabilistic inferences were derived using a
smallest worthwhile effect on performance of 1%. Substantial enhancement
of maximal endurance power output in controlled studies of subelite athletes
was very likely with artificial brief intermittent LHTL (2.6%;90%confidence
limits –1.2%), likely with LHTL (4.2%;–2.9%), possible with artificial long
continuous LHTL (1.4; –2.0%), but unclear with LHTH (0.9; –3.4%), artifi-
cial brief continuous LHTL (0.7%;–2.5%) and LLTH (0.9%;–2.4%). In elite
athletes, enhancement was possible with natural LHTL (4.0%;–3.7%), but
REVIEW ARTICLE Sports Med 2009; 39 (2): 107-127
0112-1642/09/0002-0107/$49.95/0
ª2009 Adis Data Information BV. All rights reserved.
unclear with other protocols. There was evidence that these effects were
mediated at least partly by substantial placebo, nocebo and training-camp
effects with some protocols. Enhancing protocols by appropriate manipula-
tion of study characteristics produced clear effects with all protocols
(3.5–6.8%) in subelite athletes, but only with LHTH (5.2%) and LHTL (4.3%)
in elite athletes. For .
VO
2max
, increases were very likely with LHTH (4.3%;
–2.6%) in subelite athletes, whereas in elite athletes a ‘reduction’ was possible
with LHTH (-1.5%;–2.0%); changes with other protocols were unclear.
Effects on erythropoietic and other physiological mediators provided little
additional insight into mechanisms.
In summary, natural LHTL currently provides the best protocol for
enhancing endurance performance in elite and subelite athletes, while some
artificial protocols are effective in subelite athletes. Likely mediators include
.
VO
2max
and the placebo, nocebo and training-camp effects. Modification of
the protocols presents the possibility of further enhancements, which should
be the focus of future research.
When an athlete ascends from sea level to
moderate altitude, the shortage of oxygen (hypo-
xia) initially impairs endurance training and per-
formance. After a few weeks at altitude, training
and performance recover to some extent as the
athlete adapts. If the athlete then returns to sea
level, do the adaptations lead to enhancement of
endurance performance? Coaches have long
thought so, but studies aimed at this question
appeared to be inconclusive, leading researchers
to suspect that any benefit from adaptation to
hypoxia was offset by loss of endurance fitness
consequent to the reduction in training inten-
sity.
[1]
The focus of research on altitude training
then moved from this traditional ‘live-high train-
high’ approach (LHTH) to live-high train-low
(LHTL), in which athletes live and sleep at alti-
tude, but descend regularly to lower altitude for
training sessions.
[2]
LHTL appeared to be more
successful, and interest has grown in the use of
nitrogen houses, hypobaric chambers, altitude
tents or hypoxic inhalers to adapt to hypoxia and
train normally without having to travel up and
down a mountain.
[1]
Researchers have investi-
gated three such approaches to artificial LHTL:
(i) continuous exposure to a simulated moderate
altitude for periods of 8–18 hours per day (artifi-
cial long continuous LHTL); (ii) continuous ex-
posure to a simulated moderate-high altitude for
1.5–5 hours per day (artificial brief continuous
LHTL); and (iii) intermittent exposure to a simu-
lated high altitude for <1.5 hours per day (artifi-
cial brief intermittent LHTL). The same devices
have also been used to simulate moderate altitude
while the athlete exercises continuously or inter-
mittently for at least 0.5 hours per session (artifi-
cial LLTH).
While there is general agreement that adapta-
tion to some forms of hypoxia can enhance sea-
level performance, there has been considerable
debate recently about the physiological mechan-
isms.
[3-5]
Gore and Hopkins
[4]
provided a
rationale for understanding the mechanisms un-
derlying the effects on maximal performance of
differing durations. Exercise intensities below
maximal oxygen uptake ( .
VO
2max
)[>10 minutes’
duration] are sustained essentially by aerobic
power, whereas exercise intensities above .
VO
2max
are sustained by a combination of aerobic and
anaerobic power. Aerobic power consists of three
components: (i) .
VO
2max
; (ii) the fraction of
maximal uptake that can be sustained during
the exercise; and (iii) economy or efficiency of
conversion of oxygen consumption into power
output.
[6]
Changes in endurance performance
following adaptation to hypoxia could therefore
be due to changes in any of these three compo-
nents, along with any changes in the contribution
of anaerobic power for supramaximal exercise.
Researchers who are interested in the mechanisms
108 Bonetti & Hopkins
ª2009 Adis Data Information BV. All rights reserved. Sports Med 2009; 39 (2)
underlying the performance effects of hypoxic
adaptation measure one or more of these com-
ponents or the more fundamental physiological
variables underlying them.
There has been no previous meta-analytic re-
view of the effects on performance or related
physiological measures following adaptation to
any of the artificial or natural forms of altitude
training. The current review addresses this deficit.
There were sufficient studies to allow us to meta-
analyse separately the effects on performance of
the six natural and artificial altitude protocols.
By far the most popular potential mechanism
variable has been .
VO
2max
, and we have also been
able to meta-analyse this variable with all six
protocols. Researchers have long argued that
enhancements in .
VO
2max
are mediated by erythro-
poiesis,
[1,5,7-9]
so measurements of erythro-
poietin, reticulocytes, red cell mass, haemoglobin
mass, haemoglobin concentration and ferritin
have also been reported. We have been able to
meta-analyse haemoglobin concentration for
LHTH and artificial brief intermittent LHTL, but
we had to meta-analyse haemoglobin mass and
red cell mass by combining them across all proto-
cols. We were able to perform only a graphical
analysis for erythropoietin, reticulocytes and fer-
ritin due to the small number of estimates for these
variables. Mechanisms underlying anaerobic
power are less popular with researchers, and only
the indirect measure of anaerobic power repre-
sented by peak blood lactate following an exercise
test was reported in sufficient studies for meta-
analysis in LHTH and artificial brief inter-
mittent LHTL.
1. Methodology
1.1 Study Selection
Searches of PubMed, SportDiscus and Google
Scholar were performed for studies published
in English up to and including April 2007. Re-
ference lists in review and original research arti-
cles identified were also examined. The primary
focus of the meta-analysis was performance. We
therefore included studies of performance mea-
sured at or near sea level (<1000 m). Studies
published only as conference abstracts were not
excluded. We included studies with measures of
oxygen consumption directly related to endu-
rance performance, but studies reporting haemato-
logical or other parameters not directly related to
performance and lacking a performance measure
were excluded. Several studies were excluded for
poor reporting of data or for not assessing perfor-
mance at or near sea level.
[10-25]
Other reasons for
excluding studies were: a performance enhance-
ment of 19%in 5 mmol/L lactate speed in elite
runners, when other measures of performance
increased by 0.6%and 1.1%;
[26]
the only un-
controlled study in LLTH and with only five
athletes;
[27]
and poor compliance with training, a
non-specific performance test, and an uncertain
post-exposure test time in an uncontrolled study
of the brief continuous LHTL protocol.
[28]
The
descriptive statistics for the 51 qualifying studies
are shown in table I.
1.2 Data Extraction
The study estimates for the treatment effect
were calculated for estimates without a control
group by dividing the mean post-score by the
mean pre-score for the experimental group and
expressing the ratio as a percentage; for estimates
with a control group, the post-/pre-score ratio in
the experimental group was divided by the post-/
pre-score ratio in the control group before con-
verting to a percentage. Percentage change in
performance time in time trials was converted to
change in mean power output by multiplying
by an appropriate factor derived from power-
velocity relationships.
[73]
For running, the factor
was -1; for cycling, the factor was -2.5; for
swimming, the factor was -2.0, which was an in-
dex xderived from first principles
[73]
by fitting
the power-velocity relationship P =kV
x
to pub-
lished data.
[74]
For any exercise modality, the
percentage change in time to exhaustion at a con-
stant power was converted to percentage change in
power output in an equivalent time trial of the
same duration by multiplying by a factor derived
from models for the power-duration relation-
ship of human performance, as follows: for
Performance with Adaptation to Hypoxia 109
ª2009 Adis Data Information BV. All rights reserved. Sports Med 2009; 39 (2)
Table I. Characteristics of study groups included in the meta-analysis sorted by protocol and first author
Study Subjects Sample size
a
Design Competit-
ive level
Training
phase
Hypoxic
(h/d)
b
Exposure/
intervention
days
c
Altitude
level (m)
d
Hypoxia
device
Live-high train-high
Bailey et al.
[29]
Runners 8M, 2F; 14M, 5F C Elite ? 24 28 1640
Runners 9M, 5F; 6M, 3F C Elite ? 24 28 1750
Burtscher et al.
[30]
Runners 10M; 12M C Subelite ? 24 12 2315
Friedmann et al.
[31]
Boxers +Fe
e
9M U Subelite Off-season 24 18 1800
Boxers -Fe
e
7M U Subelite Off-season 24 18 1800
Gore et al.
[32]
Cyclists 8M U Elite ? 24 31 2690
Ingjer and Myhre
[33]
Skiers 7M; 7M U Elite Competitive 24 21 1900
Jensen et al.
[34]
Rowers 9M; 9M C Elite Competitive 24 21 1822
Levine and Stray-Gundersen
[35]
Runners 10? U Subelite ? 24 28 1200
Runners 9? U Subelite ? 24 28 2500
Levine and Stray-Gundersen
[2]
Runners 9M, 4F; 9M, 4F C Subelite Competitive 24 28 2500
Miyashita et al.
[36]
Swimmers 12M, 8F U Elite Competitive 24 21 2300
Pyne
[37]
Swimmers 14M, 8F U Elite Competitive 24 21 2102
Rusko et al.
[38]
Skiers 14M; 7M C Elite ? 24 22 1700
Saunders et al.
[39]
Runners 10M; 13M C Elite ? 24 20 1750
Svedenhag and Saltinj
[40]
Runners 5M; 4M, 2F C Elite ? 24 14 2000
Svedenhag et al.
[41]
Skiers 5M, 2F U Elite ? 24 30 1900
Live-high train-low
Dehnert et al.
[42]
Triathletes 6?; 10? C Subelite ? ~18–24 13 1956/800
Levine and Stray-Gundersen
[2]
Runners 9M, 4F; 9M, 4F C Subelite Competitive ~18–24 28 2500/1200
Stray-Gundersen and Levine
[43]
Runners 6? U Subelite ? ~18–24 28 2500/1200
Stray-Gundersen et al.
[8]
Runners 8F, 14M U Elite Competitive ~18–24 27 2500/1200
Wehrlin et al.
[44]
Orienteers 5M, 5F U Elite Pre-season ~18–24 24 2456/1000
Witkowski et al.
[45]
Runners 8M, 4F U Subelite ? ~18–24 28 1780/1250
Runners 8M, 4F U Subelite ? ~18–24 28 2085/1250
Runners 8M, 4F U Subelite ? ~18–24 28 2454/1250
Runners 8M, 4F U Subelite ? ~18–24 28 2805/1250
Artificial long continuous live-high train-low
Clark et al.
[46]
Cyclists, triathletes 9M; 10M C Subelite ? 9–10 20 2650 N
2
house
Cyclists, triathletes 10M; 10M C Subelite ? 9–10 20/24 2650 N
2
house
Continued next page
110 Bonetti & Hopkins
ª2009 Adis Data Information BV. All rights reserved. Sports Med 2009; 39 (2)
Table I. Contd
Study Subjects Sample size
a
Design Competit-
ive level
Training
phase
Hypoxic
(h/d)
b
Exposure/
intervention
days
c
Altitude
level (m)
d
Hypoxia
device
Gore et al.;
[47]
Hahn et al.
[48]
Triathletes 6M; 6M C Elite ? 8–10 23 3000 N
2
house
Hahn et al.
[48]
Cyclists 5F; 6F C Elite ? 8–10 12 2650 N
2
house
Hinckson and Hopkins
[49]
Runners, triathletes 11M; 11M X Subelite ? 8 25 2500–3500 N
2
tent
Hinckson et al.
[50]
Runners 8M, 2F U Subelite ? 10 24–30/30 2500–3500 N
2
tent
Martin et al.
[51]
Cyclists 5F; 6F C Elite ? 8–10 12 2650 N
2
house
Mattila and Rusko
[52]
Cyclists 5M U Elite Competitive 18 11 3000 N
2
house
Nummela
[26]
Runners 6M, 2F; 10M C Elite ? 16–17 17 2200 N
2
house
Roberts et al.
[53]
Cyclists 14M, 5F; 14M, 5F X Subelite ? 8–10 5–15 2650 N
2
house
Rusko et al.
[38]
Skiers 9M, 3F; 8M, 2F B? ? ? 12–16 25 2500 N
2
house
Saunders et al.
[39]
Runners 10M; 13M C Elite ? 9–12 19/25 2000–3100 N
2
house
Artificial short continuous live-high train-low
Basset et al.
[54]
Skiers, skaters 7M, 5F; 7M, 5F X, B Subelite Off-season 3 6/19 3650 N
2
tent
Katayama et al.
[55]
Runners 6M; 6M C Subelite ? 1.5 9/19 4000 Chamber
Katayama et al.
[56]
Runners 8M; 7M C Subelite Competitive 3 14 4000 N
2
tent
Gore (2006); Rodriguez
(2007)
[57,58]
Swimmers 3M, 3F; 4M, 3F C Subelite ? 3–5 9 4000–5500 Chamber
Gore et al.;
[57]
Rodriguez et al.
[58]
Runners 2M, 3F; 3M, 2F C Subelite ? 3–5 9 4000–5500 Chamber
Artificial brief intermittent live-high train-low
f
Bonetti et al.
[59]
Kayakers 10M; 10M X Subelite Competitive 0.5/115/19 3600–6000 Inhaler
Bonetti et al.
[60]
Cyclists 18M; 9M C Subelite Competitive 0.5/115/19 3600–6000 Inhaler
Hamlin and Hellmans
[61]
Multisport athletes 5M, 7F; 8M, 2F C, B Subelite ? 0.75/1.5 15/19 3400–5000 Inhaler
Hinckson et al.
[62]
Rowers 2M, 5F; 1M, 4F C, B Elite ? 0.9/1.5 15/19 3600–6000 Inhaler
Julian et al.
[63]
Runners 7M; 7M C, B Elite Competitive 0.75/1.5 20/26 3600–5000 Inhaler
Wood et al.
[9]
Hockey players 15M; 14M C, B Subelite Competitive 0.6/115/19 3600–6000 Inhaler
Live-low train-high
Dufour et al.
[64]
Runners 9M; 9M C Subelite Pre-season 0.2–0.33/0.33 12/40 3000 Inhaler
Hendriksen and Meeuwsen
[65]
Triathletes 12M; 12M X, B? Subelite Pre-season 2 10 2500 Chamber
Katayama et al.
[66]
Non-athletes 7M; 7M C Trained ? 0.5 10/12 4500 Chamber
Morton and Cable
[67]
Team sports 8M; 8M C Trained ? 0.17/0.5 9/19 2750 N
2
house
Continued next page
Performance with Adaptation to Hypoxia 111
ª2009 Adis Data Information BV. All rights reserved. Sports Med 2009; 39 (2)
supramaximal tests (<7.5 minutes), the factor was
0.50/T, where T is the time in minutes;
[75]
for
submaximal tests (>7.5 minutes), the factor was
approximately 1/15.
[73]
Percentage change in time
to exhaustion in incremental tests was converted
to percentage change in peak power by multi-
plying by a factor 1-f, where f was the power of
the first stage of the test expressed as a fraction of
the peak power, under the assumption that the
load increased linearly to maximum. A spread-
sheet containing all study estimates can be ob-
tained from the authors.
1.3 Meta-Analyses
The main outcome from a meta-analysis is a
weighted mean of values of an outcome statistic
from various studies, where the weighting factor
is usually the inverse of the square of the sam-
pling standard error of the statistic. The standard
error is derived from either the confidence inter-
val or p-value of the statistic or from standard
deviations of change scores in control and ex-
perimental groups. Unfortunately, 55%of the
study-estimates for performance that would have
otherwise qualified for inclusion in our meta-
analyses did not have sufficient information to
derive the standard error; for estimates from
studies other than of intermittent artificial
LHTL, the figure is 71%. The main problem was
reporting of statistical significance or non-
significance as a p-value inequality without any
further inferential information. To exclude all
these studies from the meta-analyses would have
resulted in unacceptable bias, akin to the pub-
lication bias that arises from failure of authors to
submit studies with non-significant outcomes or
failure of journal editors to accept them. We
therefore performed the meta-analyses with a
weighting factor derived from the sample size for
each study estimate. The factor was (study sam-
ple size)/(mean study sample size). To calculate
the sample size equivalent to that of a parallel-
groups controlled trial with equal-sized groups
when the groups were of unequal size n
1
and n
2
,
we assumed equal standard error e in each group.
The standard error of the difference in means
between groups is therefore e
2
/n
1
+e
2
/n
2
, and for
Table I. Contd
Study Subjects Sample size
a
Design Competit-
ive level
Training
phase
Hypoxic
(h/d)
b
Exposure/
intervention
days
c
Altitude
level (m)
d
Hypoxia
device
Roels et al.
[68]
Cyclists, triathletes 11M; 11M C, B? Subelite Pre-season 0.2–0.5/0.5 14/47 3000 Inhaler
Terrados et al.
[69]
Cyclists 4M; 4M C Subelite ? 2 20/20–26 2300 Chamber
Truijens et al.
[70]
Swimmers 3M, 5F; 3M, 5F C, B Subelite ? 0.21/0.5 15/33 2500 Inhaler
Ventura et al.
[71]
Cyclists 6M; 5M, 1F C, B? Subelite Competitive 0.5 18/40 3200 Inhaler
a Data separated by ‘;’ are controlled trials with sample size in experimental and control groups.
b Numbers separated by ‘/’ indicate sum of time in bouts of hypoxia and sum of recovery time per session.
c Numbers after ‘/’ indicate intervention period, if longer than exposure period.
d Numbers separated by ‘/’ indicate live-high and train-low altitudes.
e Groups with and without iron supplementation.
f Altitude level estimated from arterial oxygen saturation in each study using the figure at www.high-altitude-medicine.com/SaO2-table.html.
[72]
B=blind; B? indicates blinding uncertain (assumed not blind); C=controlled trial; F=female; M=male; U=uncontrolled trial; X=crossover; ?indicates uncertain.
112 Bonetti & Hopkins
ª2009 Adis Data Information BV. All rights reserved. Sports Med 2009; 39 (2)
groups of equal size n the standard error is 2e
2
/n.
It follows that the effective sample size =2n =4
n
1
n
2
/(n
1
+n
2
). An uncontrolled trial is equivalent
to a controlled trial in which the control group
has a mean of zero and an infinite sample size, but
as n
1
-¥,4n
1
n
2
/(n
1
+n
2
)-4n
2
, so the sample
size for uncontrolled trials was inflated by a fac-
tor of 4 to make it equivalent to that of a con-
trolled trial. To ensure studies with different
numbers of estimates would have equal weight-
ing, each study’s weighting factor was divided by
the number of estimates it provided and multi-
plied by the mean number of estimates in all the
studies contributing to the meta-analysis. The
resulting meta-analysed effect is equivalent to
that produced in a random-effect meta-analysis
in which the between-study variance far out-
weighs the error variance in each study estimate,
so the confidence interval must be more con-
servative (wider) than would be provided by the
usual random-effect analysis. An assumption
underlying our analyses is that the dependent
variable giving rise to the study estimates has the
same error of measurement in all studies, but
violation of this assumption will result only in
minor differences in the weight given to each
study; the main differences in weight arise from
differences in effective sample size.
The meta-analyses were performed with the
mixed modelling procedure (Proc Mixed) in the
Statistical Analysis System (Version 9.2, SAS
Institute, Cary, NC, US). Percentage effects were
converted to factors (=1+effect/100), log trans-
formed for the analysis, then back transformed to
percentages. Study characteristics were the fixed
effects in the model; these were included as main
effects only because of the limited number of
study estimates. We limited the characteristics to
those that were included in most studies and that
might be expected on physiological or psycholo-
gical grounds to moderate the effect of hypoxia:
competitive status (elite vs subelite); design
characteristics (uncontrolled vs controlled trial,
non-blind vs blind trial); sex (males as a fraction
of the sample); training phase (competitive vs
non-competitive or unknown); altitude level or its
equivalent for artificial altitude (m); hours of
hypoxia per day (for LHTH, LHTL and artificial
long-duration LHTL); minutes of hypoxia per
day (for the remaining protocols, not counting
minutes spent in normoxia between intervals of
hypoxia); count of days when any exposure to
hypoxia occurred; total count of treatment days,
including any days resting from exposure; ratio
of exposure/treatment days; day post-exposure
when performance was tested; training intensity
on a 1–4 scale (for LLTH); type of performance
test (submaximal vs maximal); and duration of
maximal exercise tests (minutes). Missing values
for sex of nine and ten subelite runners experi-
encing LHTH
[35]
and of six subelite runners
experiencing LHTL
[43]
were assigned the mean
value of proportion of males for their protocols.
Competitive status was deemed elite if the ath-
letes were in a national team and competing
at international level. The four points of the
training-intensity scale for LLTH were: above
.
VO
2max
, 4; around .
VO
2max
, 3; around anaerobic
threshold, 2; below anaerobic threshold, 1. Post-
exposure test day and duration of maximum ex-
ercise tests were log transformed before analysis
and included as simple linear predictors. Supple-
mentary analyses (not shown in table II) were
also performed, where possible, with post-
exposure test day included as a quadratic or cubic
polynomial in ·/‚ standard deviation units, to
investigate the possibility of peaks or troughs in
performance.
An effect of a study characteristic is not shown
in the tables for one or more of the following
reasons: there was insufficient variation in the
characteristic between-study estimates to esti-
mate the effect; collinearity with other study
characteristics prevented its estimation; and the
small number of study estimates limited the anal-
ysis to only a few characteristics. To compare
effectiveness of protocols on performance, the
meta-analysed effects are shown for subelite ath-
letes (all protocols) and elite athletes (four pro-
tocols) and are adjusted to 100%controlled trials
and 100%maximal tests. For all the other study
characteristics, we could not adjust to the same
common value, so the effects on performance
for each protocol are shown evaluated at the
mean values of the study characteristics for that
protocol.
Performance with Adaptation to Hypoxia 113
ª2009 Adis Data Information BV. All rights reserved. Sports Med 2009; 39 (2)
Table II. Meta-analysis of effects on sea-level mean power output following adaptation to hypoxia experienced in studies with various protocols of natural and artificial altitude.
Effects of mean and enhanced protocols are those predicted for controlled trials and maximal tests. Effects in parentheses are unclear (>5%chance of enhancement and >5%chance
of impairment); otherwise bold indicates ‡50%chance of enhancement, italic indicates ‡50%chance of impairment, and plain font indicates ‡50%chance of trivial effect. These
probabilistic outcomes are computed with reference to a smallest important change of 1%
Effect Natural altitude protocols Artificial altitude protocols
live-high
train-high
live-high
train-low
live-high 8–18 h/d,
continuous, train-low
live-high 1.5–5h/d,
continuous, train-low
live-high <1.5 h/d,
intermittent, train-low
live-low train-
high 0.5–2h/d
Effect of mean protocol
a
(%); ––90%CL
b
Elite (1.6; –2.7) 4.0; –3.7 (0.6; –2.0) (0.2; –1.8)
Subelite (0.9; –3.4) 4.2; –2.9 1.4; –2.0 (0.7; –2.5) 2.6; –1.2 (0.9; –2.4)
Effect of enhanced protocol
c
(%); ––90%CL
Elite 5.2; –4.1 4.3; –4.1 (4.0; –5.5) (1.2; –2.5)
Subelite 4.5; –4.1 4.6; –3.3 4.8; –5.3 3.5; –3.5 3.6; –2.1 6.8; –4.9
Study characteristics changed by +1SDor-1
SD for enhanced protocol
+Altitude
-Days
exposure
+Test day
-Altitude
-Test day
+Altitude
+Hours hypoxia
-Days exposure
-Altitude
-Test day
+Exposure ratio
-Test day
-Altitude
-Train intensity
+Days exposure
+Test day
Study characteristics (mean –SD)
References 10 5 9 4 6 7
Study groups 13 9 10 5 5 7
Study estimates 33 13 17 11 33 19
Subjects/estimate 16 –712–617–915–520–617–5
Effective subjects/estimate 36 –22 41 –11 20 –915–520–617–5
Elite athletes (%)5433500330
Controlled trials (%) 46 11 85 100 100 100
Blind trials (%) 0 0 0 20 67 14
Males (%)846180728190
Competitive phase (%)313310206714
Phase unknown (%)54569060043
Maximal tests (%) 78 100 85 100 71 92
Altitude level (m) 2030 –410 2400 –290 2890 –420 4530 –840 6000 2750 –310
Hours of hypoxia per day 24 ~18–24 11 –3
Minutes of hypoxia per day 210 –84 40 –947–48
Days of exposure 23 –627–118–79–316–214–4
Continued next page
114 Bonetti & Hopkins
ª2009 Adis Data Information BV. All rights reserved. Sports Med 2009; 39 (2)
Table II. Contd
Effect Natural altitude protocols Artificial altitude protocols
live-high
train-high
live-high
train-low
live-high 8–18 h/d,
continuous, train-low
live-high 1.5–5h/d,
continuous, train-low
live-high <1.5 h/d,
intermittent, train-low
live-low train-
high 0.5–2h/d
Total period of treatment (d) 23 –627–119–714–520–330–13
Exposure/treatment ratio (%) 100 100 96 –776–32 82 –855–27
Training intensity (1–4) 2.3 –1.1
Post-exposure test day
d
9.1 ·/‚1.9 5.4 ·/‚ 2.2 2.2 ·/‚ 2.7 4.4 ·/‚ 1.9 6.4 ·/‚ 2.3 2.8 ·/‚ 1.9
Duration of maximal tests
d
(min) 6.9 ·/‚2.4 11 ·/‚ 1.3 5.6 ·/‚ 2.3 5.2 ·/‚ 3.0 6.1 ·/‚ 2.5 3.9 ·/‚ 4.9
Effects of study characteristics (%); –90%CL
Subelite-elite (0.7; –3.8) (0.3; –2.2) (0.8; –3.2) 2.4; –2.8
Uncontrolled-controlled 3.3; –3.6 -2.6; –3.0 (-1.6; –3.4)
Blind-not blind (-1.4; –4.5)
Female-male (-0.3; –3.9)
Competitive-unknown phase (0.5; –3.8)
Submaximal-maximal test (0.0; –1.6) -3.3; –2.4 (-0.3; –1.8) (1.4; –3.2)
1 SD altitude level 1.2; –1.6 -0.1; –1.0 1.5; –2.5 -2.3; –2.5 (-0.9 –2.5)
1 SD hours hypoxia 0.8; –1.8
1 SD minutes hypoxia (0.4; –2.3)
1 SD days exposure -1.8; –1.7 -1.0; –1.7 2.4; –2.5
1 SD exposure/treatment ratio 0.6; –1.2
1 SD training intensity (-1.2; –2.5)
1 SD post-exposure test day 0.5; –0.7 -0.2; –0.3 (0.1; –2.1) -0.5; –0.8 -0.4; –0.6 1.2; –1.5
1 SD duration of max. test 3.0; –2.5 -0.9; –1.2 -0.3; –0.6 0.6; –1.3 -0.2; –1.1
Random variation (%); –90%CL or ·/‚90%CL factor
Between-study SD 2.7; –2.3 1.3; –1.3 1.0; –1.9 2.2; –3.5 -0.6; –0.9 2.4; –3.1
Standard error of measurement 2.4; ·/‚1.7 0.7; ·/‚2.2 2.2; ·/‚1.9 1.2; ·/‚ 1.9 3.2; ·/‚1.3 2.8; ·/‚1.5
a Effects are the means predicted for controlled trials and maximal tests, but are otherwise evaluated at the mean values of the study characteristics for which effects are shown.
b90%CL: subtract and add this number to the observed effect to obtain the 90%CL for the true (large-sample) effect.
c Effects are the predicted means in maximal tests adjusted to –1 SD away from the mean for selected study characteristics shown.
d SD shown as ·/‚ factor derived from log-transformed times.
CL =confidence limits.
Performance with Adaptation to Hypoxia 115
ª2009 Adis Data Information BV. All rights reserved. Sports Med 2009; 39 (2)
In most models, it was possible to include a
random effect to estimate pure between-study
variation in the effect of the treatment, expressed
as a standard deviation. In principle, this measure
of between-study variation is free of sampling
variation arising from error of measurement in
the dependent variable, but use of sample size as
the weighting factor does not produce clean par-
titioning of random error into pure between-
study variation and residual error. The standard
deviation representing the residual error in such
models is the standard error of a study-estimate
with the mean sample size (n) of the meta-
analysed estimates; this standard error was multi-
plied by O(n/8) to provide an estimate of the
mean standard error of measurement of the
dependent variable. When there were insufficient
study-estimates to include a pure between-study
random effect, the residual random effect is
shown as the between-study standard deviation.
For each outcome measure, a novel funnel
plot of the inverse of an estimate’s weighting
factor (y-axis) versus the value of the estimate’s
random effect (x-axis) was examined qualita-
tively for evidence of outliers (points judged vi-
sually to be more than about 4 SDs of horizontal
scatter away from the centre of the plot) and
publication bias towards positive effects (positive
trend in the scatter). This procedure did not result
in exclusion of any estimates.
We reported uncertainty in the meta-analysed
estimates as 90%confidence limits, and we made
probabilistic magnitude-based inferences about
the true (large-sample) values of outcomes, as
described elsewhere.
[76]
In brief, an outcome was
deemed unclear if its confidence interval over-
lapped the thresholds for smallest worthwhile
positive and negative effects; equivalently, effects
were unclear if chances of the true value being
substantially positive and negative were both
>5%. The magnitude of a clear effect was re-
ported as the magnitude of its observed value,
sometimes with an estimate of the probability
the true value was substantial. The probabilities
for each meta-analysed effect and for pairwise
comparisons of effects were derived using a
published spreadsheet.
[77]
The thresholds for
smallest effects on performance were assumed to
be –1%, which is an approximate average across a
range of sports.
[78-80]
Smallest effects on .
VO
2max
and exercise economy were also assumed to be
–1%because the relationship between these
measures and endurance performance
[6]
implies
that, other things being equal, a 1%change in
either of these measures would result in a similar
change in performance. We also assumed a
smallest effect of 1%for haemoglobin or red-cell
mass because .
VO
2max
is effectively proportional
to haemoglobin mass in a cross-sectional study
of athletes.
[81]
For haemoglobin and peak lactate
concentration, there is no direct relationship
with performance; effects were therefore standar-
dized by dividing by the mean between-subject
standard deviation of these variables in the
studies that contributed to their meta-analyses,
and a modified Cohen scale was used to make
inferences.
[82]
2. Results
2.1 Exercise Performance
The meta-analysed outcomes for the six pro-
tocols of natural and artificial altitude are shown
in table II. Substantial enhancement of power
output in subelite athletes was very likely with
artificial brief intermittent LHTL, likely with
LHTL, possible with artificial long continuous
LHTL, but unclear for LHTH, artificial brief
continuous LHTL and LLTH. Comparisons be-
tween the protocols for subelite athletes revealed
that LHTL was likely better than all protocols,
with the exception of artificial brief intermittent
LHTL, where the difference was unclear. Artifi-
cial brief intermittent LHTL was possibly better
than artificial long continuous LHTL, artificial
brief continuous LHTL and LLTH. All other
differences between protocols were unclear. En-
hancements of mean power in elite athletes were
likely with LHTL, but unclear for all other pro-
tocols. In comparison with the other protocols in
elite athletes, LHTL was likely better than artifi-
cial long continuous LHTL and artificial brief
intermittent LHTL. All other differences between
protocols were unclear.
116 Bonetti & Hopkins
ª2009 Adis Data Information BV. All rights reserved. Sports Med 2009; 39 (2)
Several of the study characteristics listed in
table II moderated the effects of hypoxia; per-
formance was better in controlled relative to un-
controlled studies for LHTL, but the opposite
was observed for LHTH; subelite athletes had a
clear enhancement in performance relative to
elite athletes with artificial brief intermittent
LHTL, and submaximal exercise performance
was clearly impaired relative to maximal with
artificial long continuous LHTL. Effects for
blinding, competitive phase and sex were unclear
for the few protocols where these effects could be
estimated. Post-exposure test day had a sub-
stantial clear positive linear effect for LLTH and
trivial or unclear effects with the other protocols.
Quadratic or cubic effects of post-exposure test
day (not shown in table II) could not be modelled
with the two shortest protocols of artificial alti-
tude, and the polynomials revealed little curva-
ture with LHTL (<0.3%over ·/‚SD
2
either side
of the mean time). However, relative to the effect
at the mean post-test time, LHTH showed some
evidence of enhancement at very short times
(1.8%at ‚SD
2
or ~2.5 days; 90%confidence
limits –4.7%) followed by impairment (-1.5%at
‚SD or 5 days; –1.9%), enhancement (1.4%at
·SD or 17 days; –1.9%) and impairment (-2.3%
at ·SD
2
or 33 days; –8.9%); artificial long con-
tinuous LHTL showed a peak at the mean post-
test time with a relative impairment of 1–2%
(~–4.5%) either side of the mean (at ·/‚SD or 1
and 6 days); and LLTH showed a trough at the
mean time with relative enhancements at ‚SD or
1.5 days (3.8%;–5.3%) and at ·SD or 5 days
(1.0%;–1.4%).
The moderating effect of study characteristics
provides an avenue for enhancing each protocol,
as shown in table II for the effects on perfor-
mance after changing selected characteristics by –
or ·/‚1 SD. Improvements in power output were
observed in subelite athletes for all protocols
after these theoretical enhancements, the increase
ranging from 0.4%for LHTL to 5.9%for LLTH.
The resulting effects were all clearly beneficial for
subelite athletes, but beneficial effects for elite
athletes were clear only for LHTH and LHTL.
Alterations to the altitude level, days of expo-
sure and daily exposure hours had the biggest
contribution to the enhanced protocols, whereas
effects for other characteristics were generally
trivial or unclear. Modifying test duration by
one SD would also have produced substantial
enhancements in performance, especially for
LHTH, but this characteristic was not included
because the mean duration of tests was reason-
ably similar across the protocols, and changing
the performance test does not represent a change
to an exposure protocol.
2.2 Physiological Measures
The meta-analysed effects on sea-level .
VO
2max
are shown in table III. There was a very likely
enhancement with LHTH and a possible en-
hancement with LLTH in subelite athletes. The
trivial effect for artificial LHTL with predomi-
nantly subelite athletes is very unlikely to have
arisen from a substantial true positive effect. The
unclear effects for the remaining two artificial
protocols represent changes in .
VO
2max
that were
either unlikely to be positive (brief continuous
LHTL) for subelite athletes or were possibly po-
sitive (brief intermittent LHTL) for predomi-
nantly subelite athletes. For elite athletes, there
was a possible ‘impairment’ with LHTH, but an
unclear effect for LHTL. It was not possible to
estimate effects for elite athletes alone in the other
protocols.
Study characteristics moderating .
VO
2max
are
also shown in table III. The most interesting ef-
fect of characteristics with the natural protocols
was the increase in .
VO
2max
with increasing time
post-exposure (clear for LHTH, unclear for
LHTL), indicating that there is more benefit at
least for .
VO
2max
around 2 weeks after the inter-
vention period. The trivial effect in artificial long
continuous LHTL can be converted into a posi-
tive effect by increasing the hours of exposure;
there is also a possibility of less benefit from
‘more’ days of exposure, even though the mean
number of days of exposure is already about a
week less than for the natural protocols. A re-
duction in training intensity with LLTH would
promote a further increase in .
VO
2max
. The re-
maining effects of study characteristics on
.
VO
2max
were unclear.
Performance with Adaptation to Hypoxia 117
ª2009 Adis Data Information BV. All rights reserved. Sports Med 2009; 39 (2)
Table III. Meta-analysis of effects on sea-level maximal oxygen uptake following adaptation to hypoxia experienced in studies with various protocols of natural and artificial altitude.
Effects of mean protocol are those predicted for controlled trials. Effects in parentheses are unclear (>5%chance of increase and >5%chance of decrease); otherwise bold indicates
‡50%chance of increase, italic indicates ‡50%chance of decrease, and plain font indicates ‡50%chance of a trivial effect. These probabilistic outcomes are computed with reference
to a smallest important change of 1%
Effect Natural altitude Artificial altitude
live-high
train-high
live-high
train-low
continuous long
hypoxia (8–18 h/d),
train-low
continuous brief
hypoxia (1.5–5h/d),
train-low
intermittent brief
hypoxia (<1.5 h/d),
train-low
live-low, train-high
(0.5–2h/d)
Effect of mean protocol
a
(%); –90%CL
b
Elite -1.5; –2.0 (6.4; –11.2) -0.5; –1.4 (0.1; –2.8)
Subelite 4.3; –2.6 (6.4; –9.4) (-1.1; –3.5) 1.1; –2.0
e
Study characteristics (mean –SD)
c
References 12 5 5 4 3 8
Study groups 15 9 6 5 3 8
Study estimates 20 10 7 6 5 10
Subjects per estimate 15 –712–620–10 15 –519–516–5
Effective subje cts per estimate
d
33 –19 41 –11 20 –10 15 –519–516–5
Elite athletes (%)5733330 330
Controlled trials (%) 43 11 100 100 100 100
Blind trials (%) 0 0 0 20 33 13
Males (%) 87 61 75 72 100 91
Competitive phase (%) 29 33 0 20 100 13
Phase unknown (%) 57 56 100 60 0 50
Altitude level (m) 1990 –400 2400 –290 2680 –160 4530 –880 6000 2970 –680
Hours of hypoxia per day 10 –2
Minutes of hypoxia per day 210 –90 35 –845–46
Days of exposure 23 –627–118–69–317–314–4
Total period of treatment (d) 23 –627–119–714–521–428–14
Exposure/treatment ratio (%) 100 100 96 –976–33 78 –159–27
Training intensity (1–4) 2.1 –1.1
Post-exposure test day
c
8.0 ·/‚ 1.8 4.8 ·/‚ 2.1 1.2 ·/‚ 2.3 4.4 ·/‚ 2.0 4.7 ·/‚ 2.5 2.9 ·/‚ 2.2
Effects of study characteristics (%); –90%CL
Uncontrolled-controlled 0.3; –2.4 (-2.7; –9.3)
Competitive-unknown phase (1.3 ; –2.7)
Subelite-elite 5.5; –2.4 (-0.0; –5.2)
1 SD altitude level 0.3; –1.2 (0.0; –2.3)
Continued next page
118 Bonetti & Hopkins
ª2009 Adis Data Information BV. All rights reserved. Sports Med 2009; 39 (2)
Haemoglobin mass (including red-cell mass)
and exercise economy were meta-analysed for all
studies collectively because of the lack of study
estimates. Effects for haemoglobin mass were
unclear, but an increase in exposure days and
possibly an increase in altitude would produce a
clear increase, whereas delaying the test day
by >1SD(>10 days) would offset the increase.
The effect on exercise economy was trivial, but
a substantial increase could accrue from redu-
cing exposure days and increasing altitude
(table IV).
Haemoglobin concentration and peak lactate
could be meta-analysed only for LHTH and brief
intermittent artificial LHTL. For the interpreta-
tion of magnitude, the average pre-test between-
subject standard deviation for haemoglobin
concentration was 6.2%, while that for peak
lactate was 21%. Haemoglobin concentration
demonstrated a likely moderate increase for
LHTH and a possible small increase for artificial
brief intermittent LHTL. The moderating effect
of post-exercise test day shows that the increase
in haemoglobin concentration was lost 3–4 weeks
after exposure. The effect for peak lactate was
unclear with LHTH, but an increase in altitude
would produce a clear small to moderate de-
crease, whereas delaying the test day would pro-
duce a similar (but unexpected) decrease. Peak
lactate showed a trivial decrease for artificial
LHTL. The effect for peak lactate in artificial
brief intermittent LHTL was trivial, but the un-
certainty allows for the possibility of a small ne-
gative true effect.
Effects for other physiological measures that
could not be meta-analysed due to insufficient
data are shown in figure 1. Erythropoietin was
elevated during the hypoxic interventions and
possibly showed a small elevation afterwards.
Reticulocytes appeared to be elevated in a few
studies during the intervention. The scatter in the
plot for ferritin makes any conclusion about
trend difficult.
Plots of performance versus .
VO
2max
, haemo-
globin or red-cell mass and exercise economy are
shown in figure 2. An estimate of the strength of
the relationship between performance and each
of these variables (in units of percentage change
Table III. Contd
Effect Natural altitude Artificial altitude
live-high
train-high
live-high
train-low
continuous long
hypoxia (8–18 h/d),
train-low
continuous brief
hypoxia (1.5–5h/d),
train-low
intermittent brief
hypoxia (<1.5 h/d),
train-low
live-low, train-high
(0.5–2h/d)
1 SD hours exposure 2.5; –1.9
1 SD days exposure 0.5; –1.3 -0.9; –1.7 (1.1; –4.3) (1.0; –2.2)
1 SD training intensity -1.2; –2.1
1 SD post-exposure test day 1.0; –1.0 (1.1; –2.8) (-0.5; –3.3)
Random variation (%); –90%CL or ·/‚90%CL factor
Between-study SD 1.8; –2.4 3.8; ·/‚1.7
e
1.7; ·/‚1.9
e
3.3; ·/‚2.5
e
2.6; ·/‚2.2
e
2.1; –2.8
Standard error of measurement 2.9; ·/‚1.8 2.5; ·/‚2.6
a Effects are the means predicted for controlled trials, but otherwise evaluated at the mean values of the study characteristics for which effects are shown.
b90%CL: subtract and add this number to the observed effect to obtain the 90%CL for the true (large-sample) effect.
c SD shown as ·/‚ factor derived from log-transformed times.
d Derived by adjusting all sample sizes to those of controlled trials with equal numbers in control and experimental groups.
e Insufficient within-study clusters to estimate error of measurement; between-study SD includes within-study sampling variation.
CL =confidence limits.
Performance with Adaptation to Hypoxia 119
ª2009 Adis Data Information BV. All rights reserved. Sports Med 2009; 39 (2)
in performance per percentage change in the
variable) is provided by the slope of the regres-
sion line for each protocol (not shown in the
figure). The only clear slopes were for .
VO
2max
with LHTH (0.49 %/%;90%confidence limits
–0.29%/%) and LHTL (0.22; –0.13%/%).
Table IV. Meta-analysis of effects on sea-level haemoglobin (Hb) or red-cell mass (Hb mass), exercise economy, Hb concentration, and peak
lactate in an exercise test following adaptation to hypoxia experienced in studies with various protocols of natural and artificial altitude. Effects
in parentheses are unclear (>5%chance of increase and >5%chance of decrease); otherwise bold indicates ‡50%chance of increase, italic
indicates ‡50%chance of decrease, and plain font indicates ‡50%chance of a trivial effect. These probabilistic outcomes are computed with
reference to a smallest important change of 1%for Hb mass, 1%for economy, and 0.20 of baseline between-subject SD for Hb concentration
and peak lactate
Effect Hb mass,
where
measured
a
Economy,
where
measured
b
Hb concentration Peak lactate
LHTH intermittent
brief hypoxia,
train low
LHTH intermittent
brief hypoxia,
train low
Effect of mean protocol
c
(%);
––90%CL
d
(1.3; –2.4) 0.4; –1.3 4.8; –2.7 2.3; –1.2 (0.7; –5.7) -3.5; –4.7
Study characteristics (mean –SD)
References 12 14 5 4 5 5
Study groups 14 15 7 4 7 7
Study estimates 18 19 8 8 9 14
Subjects/estimate 15 –719–516–922–519–824–3
Effective subjects/estimate 25 –931–27 32 –11 22–535–23 24 –3
Elite athletes (%) 46335720430
Controlled trials (%) 62 80 43 100 57 100
Blind trials (%)07060050
Males (%) 749183928290
Competitive phase (%) 155329802975
Phase unknown (%) 6240430 430
Altitude level (m) 2540 –970 3410 –1460 1900 –280 6000 1990 –320 6000
Minutes of hypoxia per day 37 –735–6
Days of exposure 21 –720–624–516–222–615
Total period of treatment (d) 21 –724–624–520–422–618–2
Exposure/treatment ratio (%) 100 86 –22 100 83 –9 100 84 –9
Post-exposure test day
e
3.9 ·/‚ 2.6 3.3 ·/‚ 2.8 9.1 ·/‚ 2.1 4.3 ·/‚ 2.3 8.3 ·/‚ 2.2 5.9 ·/‚ 2.3
Effects of study characteristics (%); ––90%CL
1 SD altitude level (1.5; –2.6) 0.6; –1.6 (-1.8; –4.0
e
)-12.4; –7.0
1 SD exposure days 2.7; –2.7 -0.8 –1.6 (1.4; –7.7)
1 SD post-exposure test day -0.9; –1.0 (0.1; –1.4) -3.3; –3.9 0.6; –1.3 -10.5; –7.7 -0.6; –2.1
Random variation (%); –90%CL or ··//‚‚90%CL factor
Between-study SD 4.6; –2.2 -1.0; –2.5 3.6; ·/‚1.8
f
1.7; ·/‚1.7
f
8.7; ·/‚1.7
f
3.6; –4.8
Standard error of measurement 2.1; ·/‚1.9 6.0; ·/‚1.7 7.3; ·/‚1.5
a Number of estimates: LHTH, 10; LHTL, 3; artificial long continuous LHTL, 3; artificial brief continuous LHTL, 2.
b Number of estimates: LHTH, 4; LHTL, 3; artificial long continuous LHTL, 3; artificial brief continuous LHTL, 3 artificial brief intermittent
LHTL, 5; LLTH, 1.
c Effects are the predicted means evaluated at the mean values of the study characteristics for which effects are shown.
d90%CL: subtract and add this number to the observed effect to obtain the 90%CL for the true (large-sample) effect.
e SD shown as ·/‚ factor derived from log-transformed times.
f Insufficient within-study clusters to estimate error of measurement; between-study SD includes within-study sampling variation.
CL =confidence limits; LHTH =live-high train-high; LHTL =live-high train-low; LLTH =live-low train-high.
120 Bonetti & Hopkins
ª2009 Adis Data Information BV. All rights reserved. Sports Med 2009; 39 (2)
3. Discussion
In this first meta-analysis of sea-level exercise
performance following adaptation to hypoxic
exposure, we observed clear enhancements in
endurance power output of 1–4%in subelite
athletes with LHTL and with two of the artificial-
altitude protocols (long continuous and brief
intermittent LHTL). In elite athletes, the enhan-
cements were clear only with LHTL. Modifica-
tion of study characteristics might result in clear
enhancements of 3–7%with all protocols in sub-
elite athletes, but effects in elite athletes would be
clear only for LHTH and LHTL.
Following the development of the LHTL
approach, the use of LHTH has received little
support from sport scientists. There is enough
uncertainty in our estimates of the effect of
LHTH to allow for enhancements in elite and
subelite athletes with this protocol. Furthermore,
our estimates are for controlled trials, whereas
athletes in an altitude camp would experience the
equivalent of an uncontrolled trial, giving a pos-
sible further increase of ~3%(table II). The
LHTH protocol also showed effects of post-
exposure test day consistent with anecdotal re-
ports of coaches that performance is enhanced
immediately after altitude and peaks again sev-
eral weeks later. Taken together, these results
provide reasonable support for what is still a
widely accepted practice among many elite coa-
ches and athletes. LHTH was also one of only
two protocols that produced clear enhancements
in endurance performance for elite athletes with
appropriate manipulation of study character-
istics. These moderating effects show that it may
be better for athletes to go to higher altitudes
(~2400 m) for shorter periods (~16 days) around
2–3 weeks before an important competition.
Our results provide good evidence for the
effectiveness of LHTL, which was clearly better
than all but one protocol in subelite athletes (brief
intermittent LHTL) and elite athletes (LHTH).
The only moderating effect of study character-
istics with LHTL was unexpected: uncontrolled
trials showed a clear negative effect relative to
controlled trials. According to conventional wis-
dom, uncontrolled studies should show ‘larger’
−50
0
50
100
Effect (%)
−40
−20
0
20
40
0010 10 20
Time (d)
a
b
c
//
Hypoxia Post-hypoxia
Live-high train-high
Live-high train-low
Live-high 8−18 h/d continuous, train-low
Live high 1.5−5 h/d continuous, train-low
Live-high <1.5 h/d intermittent, train-low
Live-low, train high 0.5−2 h/d
Natural altitude:
Artificial altitude:
//
−50
0
50
100
//
Fig. 1. Individual study-estimates of effects on (a) erythropoietin,
(b) reticulocytes and (c) ferritin sampled in blood during and following
exposure to hypoxia with the various protocols of natural and artificial
altitude.
Performance with Adaptation to Hypoxia 121
ª2009 Adis Data Information BV. All rights reserved. Sports Med 2009; 39 (2)
enhancements due to the so-called ‘training-camp
effect’, which in principle is adjusted for in a
controlled trial. What may happen in reality is
that subjects in the control group of a controlled
trial experience less of the training-camp effect,
because they do not train as hard. There could
also be a contribution from a ‘nocebo effect’,
whereby subjects in a control group perform
worse, because they know they are in the control
group. There is evidence of a nocebo effect in the
classic natural LHTL study of Levine and Stray-
Gundersen
[2]
that is especially clear when the data
are presented graphically as percentage changes
(see figure 1 in Baker and Hopkins
[83]
). Indeed,
data for the effect of uncontrolled versus con-
trolled LHTL studies came entirely from this
study. Therefore, our meta-analysed effect of
~4%for controlled studies needs to be interpreted
with caution. When performance is predicted for
uncontrolled studies (as previously mentioned,
the way athletes train), the effect becomes a more
realistic ~1.5%. The only design that avoids the
nocebo problem is a blind trial, which is not
possible with natural LHTL. Further research
with controlled trials is warranted to assess the
potential of LHTL.
Artificial LHTL with long continuous expo-
sures was developed to simulate LHTL, and our
analysis provides some support for its efficacy.
The limitation with this approach appears to be
insufficient exposure to hypoxia because the
moderating effects of study characteristics show
that the effect on performance can be increased
by increasing altitude and adding daily exposure
hours. This result is consistent with the sugges-
tions of researchers who believe that at least
12 hours of daily exposure is critical for the suc-
cess of this protocol.
[7,84]
Another substantial
moderating effect was a reduction in perfor-
mance with increasing days of exposure, similar
to that with LHTH. This result for both proto-
cols seems counter-intuitive, although a ready
explanation is a short-acting placebo effect. The
only other moderating effect was a substantial
downward adjustment for submaximal perfor-
mance, which again implicates a placebo effect.
More studies are needed to clarify the role of
placebo effects with this and other protocols.
At the opposite end of the spectrum of daily
hypoxic exposure, artificial LHTL with brief in-
termittent exposures was one of the best proto-
cols in subelite athletes. The moderating effects of
study characteristics provided only marginal im-
provements of 1%, mainly through maximizing
the exposure days in the intervention period. The
equivalent altitude of this protocol is already at
the limit for ethical approval, so there is no
option to investigate higher altitudes. Alteration
−10
−5
0
5
10
ab c
−10 −50 510 −10 −50 510 −10 −50 510
Effect on maximal oxygen uptake (%)
Effect on performance (%)
Effect on Hb or red-cell mass (%) Effect on exercise economy (%)
Live-high train-high
Live-high train-low
Live-high 8−18 h/d continuous, train-low
Live-high 1.5−5 h/d continuous, train-low
Live-high <1.5 h/d intermittent, train-low
Live-low train-high 0.5−2 h/d
Natural altitude: Artificial altitude:
Fig. 2. Individual study-estimates of effects on performance plotted against maximal oxygen uptake, haemoglobin (Hb) or red-cell mass, and
exercise economy with the various protocols of natural and artificial altitude.
122 Bonetti & Hopkins
ª2009 Adis Data Information BV. All rights reserved. Sports Med 2009; 39 (2)
of the hypoxic and normoxic intervals is a possi-
ble avenue for improvement, although we have
found no clear difference between the effects of
3- and 5-minute intervals.
[60]
The clear difference
between the effect on subelite and elite athletes
suggests that the waves of hypoxia are less effec-
tive in elite athletes, possibly because elite athletes
experience more hypoxia in their muscles from
higher intensities of training compared with sub-
elite athletes.
With the remaining two forms of artificial
altitude exposure, the uncertainty in the meta-
analysed estimates was too large for their trivial
magnitudes to be clear, although clear enhance-
ments were possible with adjustment of appro-
priate study characteristics. With brief continuous
LHTL, the average altitude appears to have been
too high, since a reduction in altitude by 1 SD
could produce a substantial enhancement in per-
formance. Reducing the altitude may seem an
implausible way to enhance this protocol, but the
reduction by 1 SD would bring the altitude to
~3700 m, which is still well above that of the other
continuous protocols and which could con-
ceivably provide a sufficient hypoxic stimulus
without the negative sequelae of continuous ex-
posure to high altitude. A reduction in altitude
along with a reduction in training intensity would
also enhance performance with LLTH, but the
main enhancement for this protocol would come
from the more reasonable strategy of increasing
days of exposure. LLTH also showed evidence
that performance could be better either side of the
mean post-exposure test day (~3 days), but it
seems to us that this protocol is the least likely to
produce performance enhancement.
A study characteristic not included in the
above discussion of the individual protocols was
test duration, because altering this characteristic
does not alter the exposure protocol. There are
nevertheless implications for the effects on aero-
bic versus anaerobic performance. Our results
demonstrate that performance could be im-
proved in LHTH and brief intermittent LHTL
with tests of longer duration. In all other proto-
cols, performance could be better by a trivial
margin for shorter tests. The average test dura-
tion in all protocols was 4–11 minutes, making all
tests highly aerobic, but with only one of the
protocols (LLTH) would a 1-SD reduction in test
duration make the tests substantially anaerobic.
More studies with shorter tests are needed to
clarify the effect on anaerobic performance.
Insights into the practical application of the
findings of the meta-analyses can also be gleaned
from a consideration of the between-study stan-
dard deviations (table II). These standard devia-
tions represent unexplained variation in the mean
effect of the protocol from study to study; as
such, their magnitude is the typical deviation
from the meta-analysed mean effects that a re-
searcher or practitioner can expect to experience
in another study using the mean protocol with a
group or squad. For natural and artificial brief
intermittent LHTL, these standard deviations in
combination with the uncertainties in the mean
effects imply that most researchers and practi-
tioners will observe substantial enhancements in
performance with a group or squad of subelite
athletes. A beneficial outcome is less certain for
elite athletes with the natural protocols and for
subelite athletes with artificial long continuous
LHTL; for the remaining protocols with subelite
or elite athletes the outcomes could be good, bad
or indifferent. However, if the enhanced proto-
cols are as good as shown, the influence of the
between-study standard deviation could be nul-
lified for all protocols.
The standard errors of measurement estimated
from the meta-analyses (table II) do not have an
immediate practical application, but they do
provide evidence that the uncertainties in the
meta-analysed mean effects and in the moderat-
ing effects of study characteristics are trust-
worthy. These uncertainties are estimated from a
combination of the between-study standard de-
viations and the standard errors of measurement,
so it is important that the standard errors of
measurement estimated from the meta-analysis
are realistic. The low value for natural LHTL
(0.7%) is a reflection of the fact that almost all of
the performance tests in these studies were time
trials with runners. This value and the other va-
lues for error of measurement, given their un-
certainties, are within the normal range for tests
of endurance performance.
[73]
Performance with Adaptation to Hypoxia 123
ª2009 Adis Data Information BV. All rights reserved. Sports Med 2009; 39 (2)
It is important to understand that some in-
dividual athletes may obtain no benefit or even
impairment in performance from adaptation to
hypoxia, even with those protocols that are clearly
beneficial. Meta-analysis cannot address the ques-
tion of individual responses to treatments until re-
searchers provide complete inferential information
about experimental and control groups in the form
of confidence limits, exact p-values, or best of all,
standard deviations of change scores. Such in-
formation would also allow the use of the inverse
of sampling variance instead of sample size as a
weighting factor in the meta-analysis, which would
result in more trustworthy and probably narrower
uncertainties in the meta-analysed mean and
between-study standard deviations.
Turning to the analysis of potential mecha-
nism variables, it is clear from the findings in ta-
ble III that adaptation to hypoxia can result in
enhancements in maximal oxygen uptake. The
usual mechanism suggested for an increase in this
variable is erythropoiesis, which would effect a
change in haemoglobin or red cell mass with a
resulting increase in blood volume, cardiac output
or oxygen-carrying capacity. Our meta-analyses
provide limited evidence for this mechanism: the
meta-analysed effect on haemoglobin mass was
unclear on average, although extra exposure to
hypoxia and a higher altitude level could result in
a substantial increase. The meta-analysed effects
on haemoglobin concentration provide some ad-
ditional indirect evidence for an increase in hae-
moglobin mass, but an alternative explanation
for the increase in haemoglobin concentration
often mentioned by researchers is a dehydrating
effect of acclimatization to altitude.
[1]
Direct
evidence of erythropoiesis from levels of ery-
thropoietin and reticulocytes could not be pro-
vided by meta-analysis, due to insufficient data,
but it is reasonably clear from figure 1 that these
variables increase transiently to some extent in
some studies. Any erythropoiesis that did occur
was not accompanied by clear reductions in fer-
ritin, although supplementation with iron in most
studies would probably offset any reduction.
Do the changes in .
VO
2max
mediate the changes
in performance? The pattern of the effects on
.
VO
2max
in table III across different protocols for
elite and subelite athletes and for the moderating
effects of study characteristics does not mirror
closely the effects on performance in table II. On
the other hand, the relationships (slopes) between
changes in .
VO
2max
and performance were clear
for the natural-altitude protocols and were of a
magnitude that might be expected if .
VO
2max
was
a primary mediator, given the attenuating effects
that error of measurement in this variable would
have on the slopes. Unfortunately, blinding was
not possible with these protocols, and therefore
placebo and nocebo effects may have contributed
to the relationships. A positive relationship be-
tween changes in haemoglobin mass and changes
in performance -the expected outcome if the
changes in .
VO
2max
were mediated by erythro-
poiesis was not observed (figure 2), although
error of measurement with haemoglobin mass
(which manifests as a large between-study co-
efficient of variation, table IV) could also at-
tenuate a true substantial relationship. Thus, our
analyses have not resolved the issue of whether
.
VO
2max
is the primary mediator of performance
following adaptation to hypoxia.
There were insufficient data to meta-analyse
the effects of exercise economy for each protocol,
but a single analysis for all protocols and the
relationship between exercise economy and sea-
level performance (figure 2) provided little evidence
for this mechanism. The only other physiological
variable we meta-analysed, peak lactate con-
centration, is not a contender as a primary me-
chanism of performance enhancement, but an
increase in peak blood lactate would indirectly
implicate buffering capacity. However, placebo
and nocebo effects on performance in altitude
and control groups could also lead to an increase
in peak lactate. For the two protocols we meta-
analysed, a substantial increase in peak blood
lactate was either unlikely (LHTH) or very un-
likely (brief intermittent LHTL), so an increase in
buffering capacity is presumably not involved
with adaptation to these protocols. Other me-
chanisms therefore need to be identified, parti-
cularly for the artificial LHTL protocols, where
gains in performance appear to be due at least
partly to placebo or nocebo effects and where
an increase in .
VO
2max
may not contribute.
124 Bonetti & Hopkins
ª2009 Adis Data Information BV. All rights reserved. Sports Med 2009; 39 (2)
The suggestion of a change in cardiovascular
regulation resulting in more cardiac output to
exercising muscle
[4]
is plausible, but will be hard
to investigate.
4. Conclusions
Meta-analysis cannot adjust for the con-
founding effects of unknown or unquantified
study characteristics. Furthermore, the simplistic
nature of linear modelling, the exclusion of inter-
actions between predictors, and the inevitable
presence of substantial random error and sys-
tematic bias with some predictors all conspire to
prevent the meta-analytic model from fully ac-
counting for confounding effects even of the study
characteristics included in the model. Never-
theless, our method of estimation of confidence
intervals based on weighting by sample size is
conservative, so our analyses must provide some
evidence of the efficacy of adaptation to hypoxia
for physical performance. Subelite athletes can
experience endurance performance enhancements
with adaptation to natural altitude exposure and
to brief intermittent and long continuous proto-
cols of artificial altitude exposure. For elite ath-
letes, enhancements in endurance performance
were possible only with the natural LHTL pro-
tocol. The enhancements with natural altitude
could be mediated in part by .
VO
2max
, but placebo
effects, nocebo effects, training-camp effects and
other mechanisms may be involved with these and
the artificial protocols. Perhaps the most im-
portant outcomes of our analyses are the sugges-
tions for enhancement of the protocols, some of
which should be the focus of future research using
double-blind designs, performance measures with
smaller errors of measurement, and putative
physiological mediators. Reviewers and editors
should ensure that studies accepted for publica-
tion contain complete inferential information
about the effects in treatment and control groups.
Acknowledgements
The literature reviews in the theses of Erica Hinckson and
Matt Wood provided a valuable starting point for this review.
Chris Gore provided useful publication lists and feedback on
a draft version. The only funding for this study was provided
by our institutional employer as salaries. There are no con-
flicts of interest.
References
1. Wilber RL. Current trends in altitude training. Sports Med
2001; 31: 249-65
2. Levine BD, Stray-Gundersen J. ‘Living high-training low’:
effect of moderate-altitude acclimatization with low-
altitude training on performance. J Appl Physiol 1997; 83:
102-12
3. Gore CJ, Clark SA, Saunders PU. Nonhematological mecha-
nisms of improved sea-level performance after hypoxic
exposure. Med Sci Sports Exerc 2007; 39: 1600-9
4. Gore CJ, Hopkins WG. Counterpoint: positive effects of
intermittent hypoxia (live high-train low) on exercise per-
formance are not mediated primarily by augmented red cell
volume. J Appl Physiol 2005; 99: 2055-7
5. Levine BD, Stray-Gundersen J. Point: positive effects of
intermittent hypoxia (live high: train low) on exercise per-
formance are mediated primarily by augmented red cell
volume. J Appl Physiol 2005; 99: 2053-5
6. di Prampero PE. The energy cost of human locomotion on
land and in water. Int J Sports Med 1986; 7: 55-72
7. Levine BD. Intermittent hypoxic training: fact and fancy.
High Alt Med Biol 2002; 3: 177-93
8. Stray-Gundersen J, Chapman RF, Levine BD. ‘Living high-
training low’ altitude training improves sea level perfor-
mance in male and female elite runners. J Appl Physiol
2001; 91: 1113-20
9. Wood MR, Dowson MN, Hopkins WG. Running perfor-
mance after adaptation to acutely intermittent hypoxia.
Eur J Sport Sci 2006; 6: 163-72
10. Brugniaux JV, Schmitt L, Robach P, et al. Eighteen days
of ‘living high, training low’ stimulate erythropoiesis and
enhance aerobic performance in elite middle-distance
runners. J Appl Physiol 2006; 100: 203-11
11. Robach P, Schmitt L, Brugniaux JV, et al. Living high-
training low: effect on erythropoiesis and maximal aerobic
performance in elite Nordic skiers. Eur J Appl Physiol
2006; 97: 695-705
12. Robach P, Schmitt L, Brugniaux JV, et al. Living high-
training low: effect on erythropoiesis and aerobic perfor-
mance in highly-trained swimmers. Eur J Appl Physiol
2006; 96: 423-33
13. Schmitt L,Millet G, Robach P, et al. Influence of ‘living high-
training low’ on aerobic performance and economy of work
in elite athletes. Eur J Appl Physiol 2006; 97: 627-36
14. Adams WC, Bernauer EM, Dill DB, et al. Effects of
equivalent sea level and altitude training on .
VO
2max
and
running performance. J Appl Physiol 1975; 39: 262-6
15. Balke B, Nagle FJ, Daniels J. Altitude and maximum per-
formance in work and sports activity. JAMA 1965; 194:
646-9
16. Bushkirk ER, Kollias J, Akers RF, et al. Maximal perfor-
mance at altitude and on return from altitude in condi-
tioned runners. J Appl Physiol 1967; 23: 259-66
Performance with Adaptation to Hypoxia 125
ª2009 Adis Data Information BV. All rights reserved. Sports Med 2009; 39 (2)
17. Casas M, Casas H, Pages T, et al. Intermittent hypobaric
hypoxia induces altitude acclimation and improves the
lactate threshold. Aviat Space Environ Med 2000; 71:
125-30
18. Chapman RF, Stray-Gundersen J, Levine BD. Individual
variation in response to altitude training. J Appl Physiol
1998; 85: 1448-56
19. Daniels J, Oldridge N. The effects of alternate exposure of
altitude and sea level on world-class middle-distance run-
ners. Med Sci Sports Exerc 1970; 2: 107-12
20. Hellemans J. Intermittent hypoxic training: a pilot study.
Proceedings of the Second Annual International Altitude
Training Symposium. Flagstaff: 1999: 145-54
21. Klausen K, Robinson S, Micahel ED, et al. Effect of high
altitude on maximal working capacity. J Appl Physiol
1966; 21: 1191-4
22. Piehl Aulin K, Svedenhag J, Wide L, et al. Short-term in-
termittent normobaric hypoxia-hematological, physio-
logical and mental effects. Scand J Med Sci Sports 1998; 8:
132-7
23. Rodriguez FA, Casas H, Casas M, et al. Intermittent
hypobaric hypoxia stimulates erythropoiesis and
improves aerobic capacity. Med Sci Sports Exerc 1999; 31:
264-8
24. Rodriguez FA, Murio J, Ventura JL. Effects of intermittent
hypobaric hypoxia and altitude training on physiological
and performance parameters in swimmers [abstract]. Med
Sci Sports Exerc 2003; 35: S115
25. Telford RD, Graham KS, Sutton JR, et al. Medium altitude
training and sea level performance [abstract]. Med Sci
Sports Exerc 1996; 28: S124
26. Nummela ARI. Acclimatization to altitude and normoxic
training improve 400-m running performance at sea level.
J Sports Sci 2000; 18: 411-9
27. Vallier JM, Chateau P, Guezennec CY. Effects of physical
training in a hypobaric chamber on the physical perfor-
mance of competitive triathletes. Eur J Appl Physiol 1996;
73: 471-8
28. Rodrı
´guez FA, Ventura JL, Casas M, et al. Erythropoietin
acute reaction and haematological adaptations to short,
intermittent hypobaric hypoxia. Eur J Appl Physiol 2000;
82: 170-7
29. Bailey DM, Davies B, Romer L, et al. Implications of
moderate altitude training for sea-level endurance in elite
distance runners. Eur J Appl Physiol 1998; 78: 360-8
30. Burtscher M, Nachbauer W, Baumgartl P, et al. Benefits of
training at moderate altitude versus sea level training in
amateur runners. Eur J Appl Physiol 1996; 74: 558-63
31. Friedmann B, Jost J, Rating T, et al. Effects of iron sup-
plementation on total body hemoglobin during endurance
training at moderate altitude. Int J Sports Med 1999; 20:
78-85
32. Gore CJ, Hahn A, Rice A, et al. Altitude training at 2690m
does not increase total haemoglobin mass or sea level
.
VO
2max
in world champion track cyclists. J Sci Med Sport
1998; 1: 156-70
33. Ingjer F, Myhre K. Physiological effects of altitude train-
ing on elite male cross-country skiers. J Sports Sci 1992; 10:
37-47
34. Jensen K, Nielsen TS, Fikestrand A, et al. High-altitude
training does not increase maximal oxygen uptake or work
capacity at sea level in rowers. Scand J Med Sci Sports
1993; 3: 256-62
35. Levine BD, Stray-Gundersen J. Altitude training does not
improve running performance more than equivalent train-
ing near sea level in trained runners [abstract]. Med Sci
Sports Exerc 1992; 24: S95
36. Miyashita M, Mutoh Y, Yamamoto Y. Altitude training for
improving swimming performance at sea level. Jpn J Phys
Fitness Sports Med 1988; 37: 111-6
37. Pyne DB. Performance and physiological changes in highly
trained swimmers during altitude training. Coach Sport Sci
J 1998; 3: 42-8
38. Rusko HK, Tikkanen H, Paavolainen L, et al. Effect of
living in hypoxia and training in normoxia on sea level
.
VO
2max
and red cell mass [abstract]. Med Sci Sports Exerc
1999; 31: S86
39. Saunders PU, Telford RD, Pyne DB, et al. Improved
running economy in elite runners after 20 days of simu-
lated moderate-altitude exposure. J Appl Physiol 2004; 96:
931-7
40. Svedenhag J, Saltinj B. Aerobic and anaerobic exercise
capacities of elite middle-distance runners after two weeks
of training at moderate. Scand J Med Sci Sports 1991; 1:
205-14
41. Svedenhag J, Piehl-Aulin K, Skog C, et al. Increased left
ventricular muscle mass after long-term altitude training in
athletes. Acta Physiol Scand 1997; 161: 63-70
42. Dehnert C, Huetler M, Liu Y, et al. Erythropoiesis and
performance after two weeks of living high and training
low in well trained triathletes. Int J Sports Med 2002; 23:
561-6
43. Stray-Gundersen J, Levine BD. Altitude acclimatization
normoxic training (high/low) improves sea level endurance
immediately on descent from altitude [abstract]. Med Sci
Sports Exerc 1994; 26: S64
44. Wehrlin JP, Zuest P, Hallen J, et al. Live high-train low
for 24 days increases hemoglobin mass and red cell volume
in elite endurance athletes. J Appl Physiol 2006; 100:
1938-45
45. Witkowski S, Karlsen T, Resaland G, et al. Optimal altitude
for ‘living high–training low’ [abstract]. Med Sci Sports
Exerc 2001; 33: S292
46. Clark SA, Aughey RJ, Gore CJ, et al. Effects of live high,
train low hypoxic exposure on lactate metabolism in
trained humans. J Appl Physiol 2004; 96: 517-25
47. Gore CJ, Hahn AG, Aughey RJ, et al. Live high: train low
increases muscle buffer capacity and submaximal cycling
efficiency. Acta Physiol Scand 2001; 173: 275-86
48. Hahn AG, Gore CJ, Martin DT, et al. An evaluation of the
concept of living at moderate altitude and training at sea
level. Comp Biochem Physiol A Mol Integr Physiol 2001;
128: 777-89
49. Hinckson EA, Hopkins WG. Changes in running endurance
performance following intermittent altitude exposure si-
mulated with tents. Eur J Sport Sci 2005; 5: 15-24
50. Hinckson EA, Hopkins WG, Fleming JS, et al. Sea-level
performance in runners using altitude tents: a field study.
J Sci Med Sport 2005; 8: 451-7
126 Bonetti & Hopkins
ª2009 Adis Data Information BV. All rights reserved. Sports Med 2009; 39 (2)
51. Martin DT, Hahn AG, Lee H, et al. Effects of a 12-day ‘live
high train low’ cycling camp on 4-min and 30-min perfor-
mance [abstract]. Med Sci Sports Exerc 2002; 34: S157
52. Mattila V, Rusko H. Effect of living high and training low
on sea level performance in cyclists [abstract]. Med Sci
Sports Exerc 1996; 28: S156
53. Roberts A, Clark S, Townsend N, et al. Changes in per-
formance, maximal oxygen uptake and maximal accumu-
lated oxygen deficit after 5, 10 and 15 days of live high:
train low altitude exposure. Eur J Appl Physiol 2003; 88:
390-5
54. Basset FA, Joanisse DR, Boivin F, et al. Effects of short-
term normobaric hypoxia on haematology, muscle pheno-
types and physical performance in highly trained athletes.
Exp Physiol 2006; 91: 391-402
55. Katayama K, Matsuo H, Ishida K, et al. Intermittent hy-
poxia improves endurance performance and submaximal
exercise efficiency. High Alt Med Biol 2003; 4: 291-304
56. Katayama K, Sato K, Matsuo H, et al. Effect of intermittent
hypoxia on oxygen uptake during submaximal exercise
in endurance athletes. Eur J Appl Physiol 2004; 92:
75-83
57. Gore CJ, Rodriguez FA, Truijens MJ, et al. Increased serum
erythropoietin but not red cell production after 4 wk of
intermittent hypobaric hypoxia (4,000-5,500 m). J Appl
Physiol 2006; 101: 1386-93
58. Rodriguez FA, Truijens MJ, Townsend NE, et al. Per-
formance of runners and swimmers after four weeks of
intermittent hypobaric hypoxic exposure plus sea level
training. J Appl Physiol 2007; 103: 1523-35
59. Bonetti DL, Hopkins WG, Kilding AE. High-intensity
kayak performance after adaptation to intermittent hy-
poxia. Int J Sports Physiol Perform 2006; 1: 246-60
60. Bonetti DL, Hopkins WG, Kilding AE, et al. Cycling per-
formance following adaptation to two protocols of acutely
intermittent hypoxia. 12th Annual Congress of the Eur-
opean College of Sport Science; 2007 Jul 12-15; Jyva
¨skyla
¨
61. Hamlin MJ, Hellmans J. Effect of intermittent normobaric
hypoxic exposure at rest on hematological, physiological
and performance parameters in multi-sport athletes.
J Sports Sci 2007; 25: 431-41
62. Hinckson EA, Hopkins WG, Downey BM, et al. The effect
of intermittent hypoxic training via a hypoxic inhaler on
physiological and performance measures in rowers: a pilot
study. J Sci Med Sport 2006; 9: 177-80
63. Julian C, Gore C, Wilber R, et al. Intermittent normobaric
hypoxia does not alter performance or erythropoietic
markers in highly trained distance runners. J Appl Physiol
2004; 96: 1800-7
64. Dufour SP, Ponsot E, Zoll J, et al. Exercise training in normo-
baric hypoxia in endurance runners: I. Improvement in aerobic
performance capacity. J Appl Physiol 2006; 100: 1238-48
65. Hendriksen IJ, Meeuwsen T. The effect of intermittent
training in hypobaric hypoxia on sea-level exercise: a cross-
over study in humans. Eur J Appl Physiol 2003; 88: 396-403
66. Katayama K, Sato Y, Morotome Y, et al. Ventilatory che-
mosensitive adaptations to intermittent hypoxic exposure
with endurance training and detraining. J Appl Physiol
1999; 86: 1805-11
67. Morton JP, Cable NT. The effects of intermittent hypoxic
training on aerobic and anaerobic performance. Ergo-
nomics 2005; 48: 1535-46
68. Roels B, Millet GP, Marcoux CJ, et al. Effects of hypoxic
interval training on cycling performance. Med Sci Sports
Exerc 2005; 37: 138-46
69. Terrados N, Melichna J, Sylve
´n C, et al. Effects of training
at simulated altitude on performance and muscle metabolic
capacity in competitive road cyclists. Eur J Appl Physiol
1988; 57: 203-9
70. Truijens MJ, Toussaint HM, Dow J, et al. Effect of high-
intensity hypoxic training on sea-level swimming perfor-
mances. J Appl Physiol 2003; 94: 733-43
71. Ventura N, Hoppeler H, Seiler R, et al. The response of
trained athletes to six weeks of endurance training in hy-
poxia or normoxia. Int J Sports Med 2003; 24: 166-72
72. Hackett PH, Roach R. High-altitude medicine. In: Auer-
bach PS, editor. Wilderness medicine. 3rd ed. St Louis
(MO): Mosby, 1995: 1-37
73. Hopkins WG, Schabort EJ, Hawley JA. Reliability of power
in physical performance tests. Sports Med 2001; 31: 211-34
74. Toussaint HM, Hollander AP. Energetics of competitive
swimming: implications for training programmes. Sports
Med 1994; 18: 384-405
75. Hinckson EA, Hopkins WG. Reliability of time to exhaus-
tion analyzed with critical-power and log-log modeling.
Med Sci Sports Exerc 2005; 37: 696-701
76. Batterham AM, Hopkins WG. Making meaningful in-
ferences about magnitudes. Int J Sports Physiol Perform
2006; 1: 50-7
77. Hopkins WG. A spreadsheet for combining outcomes from
several subject groups. Sportscience 2006; 10: 51-3
78. Pyne D, Trewin C, Hopkins W. Progression and variability
of competitive performance of Olympic swimmers. J Sports
Sci 2004; 22: 613-20
79. Hopkins WG. Competitive performance of elite track-
and-field athletes: variability and smallest worthwhile
enhancements. Sportscience 2005; 9: 17-20
80. Paton CD, Hopkins WG. Variation in performance of elite
cyclists from race to race. Eur J Sport Sci 2006; 6: 1-7
81. Schmidt WF, Prommer N, Heinicke K, et al. Impact of total
hemoglobin mass on .
VO
2max
[abstract]. Med Sci Sports
Exerc 2007; 39 (5): S3
82. Snowling NJ, Hopkins WG. Effects of different modes of
exercise training on glucose control and risk factors for
complications in type 2 diabetic patients: a meta-analysis.
Diabetes Care 2006; 29: 2518-27
83. Baker A, Hopkins WG. Altitude training for sea-level
competition. Sportscience 1998; 1 [online]. Available
from URL: http://sportsci.org/traintech/altitude/wgh.html.
[Accessed 2009 Jan 18]
84. Rusko HK, Tikkanen HO, Peltonen JE. Altitude and
endurance training. J Sports Sci 2004; 22: 928-45
Correspondence: Prof. Will G. Hopkins, School of Sport and
Recreation, AUT University, Private Bag 92006, Auckland
0627, New Zealand.
E-mail: will@clear.net.nz
Performance with Adaptation to Hypoxia 127
ª2009 Adis Data Information BV. All rights reserved. Sports Med 2009; 39 (2)
