British Journal of Anaesthesia 113 (S1): i74i81 (2014)

Advance Access publication 23 May 2014 . doi:10.1093/bja/aeu110

Risk of new or recurrent cancer after a high perioperative

inspiratory oxygen fraction during abdominal surgery
C. S. Meyhoff 1*, L. N. Jorgensen 2, J. Wetterslev 3, V. D. Siersma 5 and L. S. Rasmussen 4 PROXI Trial Group

1
Department of Anaesthesiology, Herlev Hospital, University of Copenhagen, Herlev Ringvej 75, DK-2730 Herlev, Denmark
2
Digestive Disease Center, Bispebjerg Hospital, University of Copenhagen, Bispebjerg Bakke 23, DK-2400 Copenhagen NV, Denmark
3
Copenhagen Trial Unit, Centre for Clinical Intervention Research and 4 Department of Anaesthesia, Centre of Head and Orthopaedics,
Rigshospitalet, University of Copenhagen, Blegdamsvej 9, DK-2100 Copenhagen , Denmark
5
The Research Unit for General Practice and Section of General Practice, Department of Public Health, University of Copenhagen, ster
Farimagsgade 5, DK-1014 Copenhagen K, Denmark
* Corresponding author. E-mail: christianmeyhoff@gmail.com

Background. Administration of supplemental oxygen in the perioperative period is

Editors key points
High inspired oxygen
supplementation during
anaesthesia may be
associated with adverse
outcomes.
This post hoc analysis of
the PROXI trial evaluated
the association between
F IO2 = 0.8 or F IO2 = 0.3 and
cancer incidence or
recurrence after a median
time of 4 yr.
Although the incidence of
cancer was similar, higher
mortality was observed in
the high inspired O2 group,
and shorter cancer-free
survival.
controversial, as it may increase long-term mortality. Our aim was to assess the association
between 80% oxygen and occurrence of subsequent cancer in patients undergoing
abdominal surgery in a post hoc analysis of the PROXI trial.
Methods. The 1386 patients in the PROXI trial underwent elective or emergency laparotomy
between 2006 and 2008 with randomization to either 80% or 30% oxygen during and for
2 h after surgery. We retrieved follow-up status regarding vital status, new cancer
diagnoses, and new histological cancer specimens. Data were analysed using the Cox
proportional hazards model.
Results. Follow-up was complete in 1377 patients (99%) after a median of 3.9 yr. The primary
outcome of new cancer diagnosis or new malignant histological specimen occurred in 140 of
678 patients (21%) in the 80% oxygen group vs 150 of 699 patients (21%) assigned to 30%
oxygen; hazards ratio 1.06 [95% confidence interval (CI) 0.84, 1.34], P0.62. Cancer-free
survival was significantly shorter in the 80% oxygen group; hazards ratio 1.19 (95% CI 1.01,
1.42), P0.04, as was the time between surgery and new cancer, median 335 vs 434 days in
the 30% oxygen group. In patients with localized disease, non-significant differences in
cancer and cancer-free survival were found with hazard ratios of 1.31 and 1.29, respectively.
Conclusions. Although new cancers occurred at similar rate, the cancer-free survival was
significantly shorter in the 80% oxygen group, but this did not appear to explain the excess
mortality in the 80% oxygen group.
Clinical trial registration. ClinicalTrials.gov (NCT01723280).
Keywords: laparotomy; neoplasms; oxygen therapy; postoperative complications
Accepted for publication: 24 February 2014

Oxygen is essential for human survival, but the oxygen partial
pressure is drastically reduced to the site of utilization,1 2
because formation of reactive oxygen species (ROS) may
damage cell components including DNA if the ROS are not neu-
tralized by antioxidant systems.3 4
An overload of oxygen may thus not be desirable,4 and
recent studies have indicated increased morbidity and mortal-
ity with excess oxygen administration in various emergency
medical conditions.5 8 On the other hand, oxygen supple-
mentation is often used during general anaesthesia, where
an impairment of cardiovascular and pulmonal function can
compromise oxygen delivery.9 A high perioperative inspiratory
oxygen fraction (F IO2 ) of 0.80 has therefore been proposed to
prevent wound infection, and some randomized trials have
shown benefit,10 12 whereas others, including the PROXI
trial, have not.13 16
In the PROXI trial, 30 day mortality tended to be higher
when 80% oxygen was given during laparotomy compared
with 30% oxygen, and a subsequent analysis detected in-
creased long-term mortality with hyperoxia, predominantly
in patients undergoing cancer surgery.17 The effect of hyper-
oxia on cancer development is, however, sparsely elucidated,
although high levels of oxygen in addition to DNA damage
induce angiogenesis and may increase erythropoietin produc-
tion, which is a potent cellular growth factor.3 18 19 Angiogen-
esis, erythropoietin, and DNA damage from oxidative stress

& The Author [2014]. Published by Oxford University Press on behalf of the British Journal of Anaesthesia. All rights reserved.
For Permissions, please email: journals.permissions@oup.com
80% oxygen and risk of subsequent cancer
could thus lead to growth of new or remaining tumour cells
after otherwise curative surgery. Thus, we found it important
to evaluate the risk of developing new or subsequent cancers
after exposure to perioperative hyperoxia.
The aim of this post hoc study of the PROXI trial13 was
to assess the possible association between an F IO2 of 0.80
during and after laparotomy on the occurrence of a subse-
quent, new or recurrent, cancer over a 4 yr follow-up period.
Methods
Approval was obtained for the PROXI trial and this follow-up
study from the regional Ethics Committee (The Capital
Region, Denmark, KF 02 306766), the Danish Medicines
Agency (registered with EudraCT no. 2006-001710-32), and
the Danish Data Protection Agency.13 20 After written informed
consent, 1386 patients were included in 14 Danish hospitals
between October 8, 2006, and October 6, 2008. Patients were
aged 18 yr or older, and undergoing elective or emergency
laparotomy. Exclusion criteria were: surgery performed under
general anaesthesia within the preceding 30 days, chemo-
therapy for malignant disease within the preceding 3
months, inability to give informed consent, and preoperative
arterial oxygen saturation below 90% without supplemental
oxygen.13 20
Patients were randomized to either the 80% or the 30%
oxygen group with stratification for: centre, diabetes mellitus,
elective or emergency surgery, and body mass index (BMI)
(,30 or 30 kg m22).20 A trial protocol was used to standard-
ize key elements of perioperative care, including temperature
and glucose control, prophylactic antibiotics administration,
epidural analgesia, general anaesthesia without nitrous
oxide, and normovolaemic fluid therapy.20
The interventions consisted of an F IO2 of 0.80 or 0.30 given
after tracheal intubation and until tracheal extubation. For
the first 2 postoperative hours, the allocated oxygen concen-
tration was delivered through a non-rebreathing face mask
with reservoir with a flow of either 14 litres of O2 and 2 litres
of air per minute or 2 litres of O2 and 14 litres of air per
minute. The intervention was blinded to patients, surgeons,
the staff on the wards, outcome assessors, and statisticians.20
Extended follow-up was performed on September 28, 2011,
where data were collected from the Danish National Patient
Registry, the Danish Cancer Registry, the Danish Cause of
Death Registry, and the Danish Patobank (collecting all histo-
logical specimens in Denmark). There are mandatory registra-
tion to all databases from all Danish hospitals, and all Danish
citizens are registered with a unique 10 digit number.21
Cancer diagnoses were classified according to the ICD-10
and graded with the tumour, node, and metastasis classifica-
tion (TNM), and underwent extensive validation at the Danish
Cancer Registry for duplicate information and consistency. In
cases with multiple previous cancers, we classified T, N, and
M as the worst grade, respectively. Metastatic spread was
defined as any N.0 or any M1. Patients with a cancer diagno-
sis, but unknown N and M status were classified as having no
metastatic spread.
BJA
We classified all patients into the following groups at
baseline:
No cancer, defined as no cancer diagnosed before ran-
domization.
Localized cancer, defined as any cancer without meta-
static spread diagnosed before randomization or within
30 days after surgery, if the cancer was diagnosed
during index surgery.
Metastasized cancer, defined as any cancer with meta-
static spread diagnosed before randomization or within
30 days after surgery, if the cancer was diagnosed
during index surgery.
Outcomes
The primary outcome measure was a composite outcome
measure of either a new or subsequent cancer diagnosis or a
new or subsequent histological cancer specimen. Cancer
diagnosis was defined as any cancer diagnosis in the Danish
Cancer Registry after index surgery, but excluding those
cases of cancer that were diagnosed within 30 days after
index surgery. Recurrence of a cancer is not reportable to the
Danish Cancer Registry, and may therefore only be detected
in histological specimens (such as lung metastases without
specification), which therefore were included in the primary
outcome. Histological specimens from the Danish Patobank
were analysed according to the Systematized NOmenclature
of MEDicine (SNOMED) code, and a histological cancer specimen
was defined as any specimen indicating a new or recurrent
malignant neoplasm between 31 days after index surgery
and follow-up. The Danish Cancer Registry has a 12 yr delay
because of extensive data validation, and follow-up in this
registry is therefore December 31, 2009, but follow-up for the
primary outcome was September 28, 2011, where any indica-
tion of cancer in the two registries was retrieved.
Secondary outcomes included: new or subsequent cancer
diagnosis, any new histological cancer specimen, new cancer
specimen of a histological type not previously found, cancer-
free survival, and readmissions with cancer as primary or sec-
ondary diagnosis code. Readmissions were defined as any
new admission to a Danish hospital after discharge from
index surgery. If the postoperative hospital course involved
admissions to other wards, intensive care units, or other hospi-
tals, we defined readmission as the first hospital admission
after final discharge. Diagnosis codes for readmission were
classified according to the primary ICD-10 and the secondary
ICD-10 codes. Cancer-free survival was the time from index
surgery to the first of the following events: death, new can-
cer diagnosis, or new histological specimen showing any
malignant neoplasm within the longest follow-up period
until September 28, 2011. As other explorative outcomes, we
described histological type of new cancer specimen and
cause of death, which was classified according to the primary
and secondary ICD-10 codes for cause of death listed on the
death certificate. The Danish Cause of Death Registry has a
12 yr delay for data validation, and the follow-up date for
cause of death was therefore December 31, 2009.
i75
BJA
Statistical analysis
The statistical analyses were planned before data were avail-
able (NCT01723280). The primary and secondary outcomes
were first investigated in the KaplanMeier plots of the survival
curves, stratified for 80% vs 30% oxygen and additionally on
cancer status at the time of index surgery. Thereafter, the
primary and secondary outcomes were analysed in the Cox
proportional hazards models in which a hazard ratio (HR) mea-
sured the independent effect of 80% vs 30% oxygen. These
analyses were done both unadjusted and adjusted for the
following design and stratification variables:13 22 elective or
emergency surgery, diabetes mellitus, BMI (,30 or 30 kg m22),
trial site, daily smoking, age (,40 or 40 yr), alcohol con-
sumption .48 g day21, ASA physical status score, cardiovascu-
lar disease other than hypertension, perioperative blood
transfusion, and cancer status at the time of index surgery.
The adjusted HR was used in the primary analyses. Time
between surgery and events was compared using the Wilcoxon
rank-sum test.
Subgroup analyses were performed according to patients
with no cancer, localized cancer, or metastatic cancer at base-
line. Additionally, two sensitivity analyses were planned: one in
which follow-up for the primary outcome was December 31,
2009, where complete data sets from the Danish Cancer
Registry and the Danish Patobank were available, and
another which evaluated the primary outcome across sub-
groups when patients with unknown status for disseminated
cancer disease were classified as having metastasized cancer.
The original sample size calculation is described else-
where.20 All intervention effect estimates were analysed
according to the intention-to-treat principle and reported
with 95% confidence interval (CI) and a two-sided P,0.05
was considered statistically significant. Analyses were per-
formed using SAS for Windows, version 9.3 (SAS Institute,
Inc., Cary, NC, USA).
Results
The PROXI trial included 1386 patients, and at follow-up, six
patients were no longer Danish citizens, and three patients
could not be located in the patient registries. Complete data
were obtained from 1377 patients (99.4%) after a median
follow-up of 3.9 (range 2.95.0) yr with comparable patient
characteristics at baseline in the two groups (Table 1).
There were 57 new cancer diagnoses and 782 histological
specimens in 290 patients after index surgery showing malig-
nant neoplasia. The primary composite outcome of either a
subsequent cancer diagnosis or a new histological cancer spe-
cimen occurred in 140 of 678 patients (21%) in the 80% oxygen
group vs 150 of 699 patients (21%) patients assigned to 30%
oxygen; HR 1.06 (95% CI 0.84, 1.34), P0.62 (Table 2). There
were no significant differences in new histological malignancy
or readmission for cancer (Tables 2 and 3). Long-term mortality
was significantly higher in the 80% oxygen group, with an HR
of 1.29 (95% CI 1.05, 1.58), P0.02 (Table 2), and cancer-free
survival was lower in the 80% oxygen group with an HR for
death or cancer of 1.19 (95% CI 1.01, 1.42), P0.04 (Fig. 1,
i76
Meyhoff et al.
Table 2). In patients meeting the composite outcome, the
time between surgery and new cancer was shorter in 80%
oxygen group, median 335 vs 434 days in the 30% oxygen
group, P0.047 (Table 4).
In the subgroup analysis according to baseline cancer
status, the largest differences were found in patients with loca-
lized cancer. The median time between surgery and the
primary outcome of new cancer was 390 and 567 days
(P0.03) in the 80% and 30% oxygen group, respectively.
Cancer-free survival was not significantly different in patients
with localized cancer; HR for death or cancer 1.29 (95% CI
0.97, 1.72), P0.08 (Fig. 2, Table 2).
In the first sensitivity analysis, where follow-up was the
latest time with available data sets in the Danish Patobank
and the Danish Cancer Registry, the primary outcome was ful-
filled in 118 of 678 patients (17%) vs 109 of 699 patients (16%)
in the 80% and 30% oxygen group, respectively, HR 1.17 (95%
CI 0.89, 1.52), P0.26. Cancer-free survival was shorter in the
80% oxygen group, HR for death or cancer 1.24 (95% CI 1.02,
1.50), P0.03.
The second sensitivity analysis, in which unknown baseline
cancer dissemination was classified as metastatic disease,
showed a slightly larger difference in the primary outcome 47
of 175 patients (27%) vs 37 of 181 patients (20%), HR 1.56
(95% CI 1.01, 2.41), P0.04, and a slightly larger difference in
cancer-free survival, HR for death, or cancer 1.43 (95% CI
1.04, 1.98), P0.03, when comparing 80% with 30% oxygen
in patients with localized disease.
As explanatory analysis, we performed a survival analysis
with new cancer as censoring event in order to investigate
whether a higher cancer occurrence may explain the previously
found higher mortality (Table 2). This analysis showed a slightly
more pronounced risk of long-term mortality; HR 1.40 (95% CI
1.08, 1.80).
Discussion
In this long-term follow-up study of new or recurrent cancers
after 80% vs 30% oxygen was given during abdominal
surgery, we found that the overall frequency of cancers were
similar. However, the time to new cancer and cancer-free sur-
vival was shorter in the 80% oxygen group, with the largest
burden of early cancer and death among patients with loca-
lized disease at baseline.
This first description of association between supplemental
oxygen and subsequent cancers with almost complete follow-
up rate (99.4%) of a randomized trial are the primary strengths
of our study together with the high validity of data in the na-
tional patient registries in Denmark,21 which receive manda-
tory reporting from all hospitals based on a unique ID for
every patient.
The extended follow-up of our trial comes with limitations
as well. The outcomes were defined after enrolment of the
1400 patients in the trial, a sample size calculated to evaluate
surgical site infection. On the other hand, we defined and pub-
lished the research plan for this follow-up study before cancer
data were extracted. We had 86% power to detect an HR of
80% oxygen and risk of subsequent cancer BJA

Table 1 Characteristics of 1377 patients undergoing laparotomy. Values are presented as median (5 95% range). All characteristics have ,0.5%
missing data, unless stated below. *Previous cancer defined as a cancer diagnosed before index surgery. N status was unknown for 27 vs 29
patients, and M status was unknown for 20 vs 25 patients in the 80% and 30% oxygen group, respectively. Current cancer at index surgery defined
as a cancer diagnosed in connection to index surgery. Tstatus was unknown for 38 vs 38 patients, N status was unknown for 48 vs 52 patients, and M
status was unknown for 28 vs 35 patients in the 80% and 30% oxygen group, respectively. Localized cancer defined as previous or current cancer at
index surgery with N0 and M0; metastasized cancer defined as either previous or current cancer at index surgery with N.0 or M1

Characteristic 80% oxygen group (n5678) 30% oxygen group (n5699)

Age (yr) 64 (27 85) 64 (34 84)
Gender (M/F) 284/394 292/407
BMI (kg m22) 25 (18 35) 25 (19 35)
ASA physical status 174/371/129/4 196/373/125/5
class (I/II/III/IV)
Emergency surgery 185 (27%) 194 (28%)
History of
Current smoking 202 (30%) 211 (30%)
Chronic obstructive pulmonary disease 35 (5.2%) 34 (4.9%)
Other respiratory disease 45 (6.6%) 48 (6.9%)
Alcohol consumption .48 g day21 28 (4.1%) 35 (5.0%)
Diabetes mellitus 51 (7.5%) 53 (7.6%)
Hypertension 207 (31%) 186 (27%)
Other cardiovascular disease 124 (18%) 96 (14%)
Epidural analgesia 468 (69%) 498 (71%)
Surgical procedure
Colorectal procedures 301 (44%) 329 (47%)
Other surgery 377 (56%) 370 (53%)
Duration of surgery (min) 126 [38 310] 132 [35 295]
Peroperative blood transfusion 121 (18%) 120 (17%)
Previous cancer status*
Any previous cancer 137 (20%) 157 (22%)
Any previous N.0 28 (4.1%) 33 (4.7%)
Any previous M1 14 (2.1%) 13 (1.9%)
Current cancer at index surgery
Any current cancer at index surgery 333 (49%) 348 (50%)
T0 2 (0.3%) 1 (0.1%)
T1 51 (7.5%) 50 (7.2%)
T2 33 (4.9%) 32 (4.6%)
T3 128 (19%) 135 (19%)
T4 81 (12%) 92 (13%)
N0 148 (22%) 159 (23%)
N1 74 (11%) 87 (12%)
N2 54 (8.0%) 45 (6.4%)
N3 9 (1.3%) 5 (0.7%)
M0 242 (36%) 230 (33%)
M1 63 (9.3) 83 (12%)
Histology of previous or current cancer
Adenocarcinoma 326 (48%) 353 (51%)
Basocellular carcinoma 28 (4.1%) 28 (4.0%)
Planocellular carcinoma 14 (2.1%) 9 (1.3%)
Other carcinoma 55 (8.1%) 56 (8.0%)
Haematological malignancy 16 (2.4%) 9 (1.3%)
Other malignant neoplasia 68 (10%) 62 (8.8%)
Cancer status classification
No cancer 264 (39%) 260 (37%)
Localized cancer 208 (31%) 221 (32%)
Metastasized cancer 206 (30%) 218 (31%)

i77
BJA Meyhoff et al.

Table 2 Long-term outcomes after laparotomy. Percentages are calculated per group or subgroup where relevant. HRs are presented with the 30%
oxygen group as reference group. P-values are calculated with the Cox proportional hazards ratio. *Adjusted for elective or emergency surgery,
diabetes mellitus, BMI (,30 or 30 kg m22), trial site, daily smoking, age (,40 or 40 yr), alcohol consumption .48 g day21, ASA physical status
score, cardiovascular disease other than hypertension, perioperative blood transfusion, and cancer status at the time of index surgery

Outcome 80% oxygen 30% oxygen Adjusted hazards P-value Unadjusted hazards P-value
group (n5678) group (n5699) ratio (95% CI)* ratio (95% CI)
Primary outcome, new or subsequent cancer
diagnosis or new histological cancer specimen
Overall 144 (21%) 150 (21%) 1.06 (0.84 1.34) 0.62 1.03 (0.82 1.29) 0.81
No cancer (n264 vs 260) 16 (6.0%) 18 (6.9%) 0.93 (0.47 1.83) 0.84 0.87 (0.45 1.71) 0.70
Localized cancer (n208 vs 221) 57 (27%) 54 (24%) 1.31 (0.90 1.91) 0.16 1.25 (0.86 1.81) 0.24
Metastasized cancer (n206 vs 218) 71 (34%) 78 (36%) 0.93 (0.67 1.29) 0.65 1.02 (0.74 1.41) 0.89
New or subsequent cancer diagnosis 25 (3.7%) 30 (4.3%) 0.88 (0.51 1.50) 0.63 0.88 (0.52 1.50) 0.65
New histological cancer specimen 141 (21%) 144 (21%) 1.08 (0.85 1.36) 0.53 1.05 (0.83 1.32) 0.68
New cancer specimen of a histological type 33 (4.9%) 37 (5.3%) 0.95 (0.59 1.52) 0.82 0.95 (0.59 1.51) 0.81
not previously found
Readmission with cancer as primary 264 (39%) 270 (39%) 1.01 (0.85 1.21) 0.87 1.02 (0.86 1.20) 0.85
diagnosis
Readmission with cancer as secondary 295 (44%) 295 (42%) 1.06 (0.89 1.26) 0.53 1.03 (0.87 1.22) 0.72
diagnosis
Mortality
Overall 208 (31%) 182 (26%) 1.29 (1.05 1.58) 0.02 1.21 (0.99 1.48) 0.06
No cancer (n264 vs 260) 26 (10%) 21 (8.1%) 1.24 (0.69 2.21) 0.47 1.23 (0.69 2.18) 0.49
Localized cancer (n208 vs 221) 70 (34%) 57 (26%) 1.28 (0.90 1.82) 0.18 1.38 (0.98 1.96) 0.07
Metastasized cancer (n206 vs 218) 112 (54%) 104 (48%) 1.30 (0.99 1.72) 0.06 1.19 (0.91 1.56) 0.19
Mortality or cancer
Overall 280 (41%) 268 (38%) 1.19 (1.01 1.42) 0.04 1.12 (0.95 1.32) 0.19
No cancer (n264 vs 260) 36 (14%) 34 (13%) 1.08 (0.67 1.73) 0.75 1.04 (0.65 1.66) 0.86
Localized cancer (n208 vs 221) 102 (49%) 90 (41%) 1.29 (0.97 1.72) 0.08 1.33 (1.00 1.77) 0.048
Metastasized cancer (n206 vs 218) 142 (69%) 144 (66%) 1.16 (0.91 1.47) 0.23 1.10 (0.88 1.39) 0.40
Mortality with cancer as censoring event
Overall 136 (20%) 118 (17%) 1.40 (1.08 1.80) 0.01 1.23 (0.96 1.57) 0.10
No cancer (n264 vs 260) 20 (7.6%) 16 (6.2%) 1.22 (0.63 2.37) 0.56 1.23 (0.64 2.38) 0.53
Localized cancer (n208 vs 221) 45 (22%) 36 (16%) 1.33 (0.85 2.07) 0.21 1.46 (0.94 2.26) 0.09
Metastasized cancer (n206 vs 218) 71 (35%) 66 (30%) 1.50 (1.05 2.13) 0.03 1.20 (0.86 1.86) 0.29

1.19 for cancer-free survival with 5% type 1 error risk based on
our subjects included over 730 days with additional follow-up
for 1087 days after inclusion of last patient, the median cancer-
free survival time being 363 days in the control group, but the
power to detect an HR of 1.06 for our primary outcome is
much lower. Although an HR of 1.19 for cancer-free survival
must be considered important, smaller differences are also
clinically relevant, and the power to explore this in the sub-
groups is even smaller. Secondly, we could not measure the dis-
tribution between recurrent cancers and new cancers among
our primary outcome, because information about peroperative
margin-free resection was not available, but our composite
primary outcome is the most sensitive measure of new
cancer burden for the patient. The sensitivity analysis with
the shortest follow-up did in fact result in marginally larger dif-
ferences between the 80% and 30% oxygen group. Thirdly, the
patient population underwent a broad range of elective and
emergency laparotomy procedures, and only some of the
procedures involved manipulation of a tumour, which could in-
fluence the risk of micrometastases and cancer recurrence.23
24
On the other hand, although tumour stage was not available,
we had thorough knowledge of all known previous and current
cancers, and those were evenly distributed between the
groups.
We observed a virtually identical frequency of new cancers
in the 80% and 30% oxygen group, but cancer-free survival
and time to cancer were shorter in the 80% oxygen group, rec-
ognizing that time to new cancer is an explanatory outcome
and a much lower P-value is needed to achieve firm evidence.25
We have previously reported increased long-term mortality in
our 80% oxygen group,17 and it is therefore likely that the
problem of competing outcomes could influence the interpret-
ation of certain results, that is: the patients in the 30% oxygen
group live longer and have thus a higher risk of developing a
new cancer in the follow-up period, which again can mask a po-
tentially increased frequency of cancers in the 80% oxygen

i78
80% oxygen and risk of subsequent cancer BJA
Nevertheless, we found that patients in the localized sub-
Table 3 New histological findings and cause of death after group, that is, patients with previous or current cancer at the
laparotomy. *Cause of death is reported until follow-up of the
time of index surgery who has no lymph node or distant metas-
Danish Cause of Death Registry, December 31, 2009
tases, may be somewhat vulnerable to perioperative hyper-
Outcome 80% oxygen 30% oxygen oxia. Subsequent cancers in this subgroup represent
group (n5678) group (n5699) recurrence or a new cancer, and we found statistically non-
Histology of new cancer specimen significant hazards ratios of 1.31 for new cancer and 1.29 for
Adenocarcinoma 84 (12%) 90 (13%) cancer or death, and these differences were in fact statistically
Basocellular carcinoma 20 (3.0%) 24 (3.4%) significant in the sensitivity analysis with unknown N and M
Planocellular carcinoma 4 (0.6%) 2 (0.3%) status classified as metastatic disease. Further, if new cancer
Other carcinoma 23 (3.4%) 23 (3.3%) developed in the localized subgroup, this occurred at a
Haematological 6 (0.9%) 3 (0.4%) median of 177 days earlier in the 80% oxygen group.
malignancy A number of aspects are important during surgery for the
Other malignant neoplasia 23 (3.4%) 31 (4.4%) course of a localized cancer, because manipulation of the
Cause of death* tumour may result in release of tumour cells to the blood or
All-cause mortality 152 (22%) 127 (18%) lymph stream24 and micrometastases may be present in one-
Cancer 107 (70%) 95 (75%) quarter of lymph node-negative patients.23 The progression of
Cardiovascular 14 (9.2%) 8 (6.3%) such minimal residual disease is dependent on intact immune
Infection 10 (6.6%) 7 (5.5%) function, and a number of perioperative factors are in fact
Other 10 (6.6%) 6 (4.7%) depressing the cell-mediated immune system, and increasing
Unknown 11 (7.2%) 11 (8.7%) angiogenesis and growth factors that promote local and distant
growth of malignant cells.24 Supplemental perioperative oxygen
has a strong positive effect on angiogenesis,18 whereas the
effects on growth factor stimulation and the local immune
system are not fully clarified,19 26 leading to hypotheses with
tumour suppression and tumour progression.19 Hyperoxia may
100
increase the production of erythropoietin, a potent cellular
growth factor,9 27 and the formation of ROS in the hyperoxic en-
80 vironment may promote DNA damage.3 19 26 29 The sum of
% cancer-free survival

angiogenesis, erythropoietin production, and oxidative stress

60
40
20
30% oxygen
0 80% oxygen
0 1 2 3
Years since operation
Fig 1 Cancer-free survival after abdominal surgery.
group. Therefore, to alleviate this problem, we chose to analyse
the primary and secondary outcomes in a model for
time-to-event datathe Cox regression model.
Some of our findings oppose cancer as causing the
increased mortality, because the point estimate of HR of time
to death with cancer as censoring event (1.40) was in fact
slightly higher than the HR of time to death (1.29). This sug-
gests that other explanations than new cancer should
explain the observed increased mortality in the 80% oxygen
group. We could furthermore not identify a specific type of hist-
ology that was more vulnerable to perioperative hyperoxia.
could thus lead to growth of minimal residual disease after
otherwise curative surgery.
Other studies have investigated long-term outcomes after
anaesthesia, and a 3.5 yr follow-up of the ENIGMA trial,30
comparing 70% nitrous oxide with 80% oxygen, found no sig-
nificant difference in long-term mortality. The investigated
interventions may, however, have resulted in similar peri-
operative harm. The frequency of new cancers and can-
cer-free survival in our study are comparable with those
4 5 found in studies on epidural analgesia to prevent cancer
recurrence.31 32
Recently, major concerns have been raised with hyperoxia in
emergency medical settings.4 A randomized trial of patients
with exacerbation of chronic obstructive pulmonary disease
found increased mortality with prehospital hyperoxia.5 Large
retrospective studies have found higher morbidity and mortal-
ity when hyperoxia is present after resuscitation from cardiac
arrest and traumatic brain injury,6 7 and a Cochrane review
suggest increased mortality after myocardial infarction with
supplemental oxygen therapy.8
Our study can generate new hypotheses, but not prove caus-
ality between perioperative hyperoxia, subsequent cancers, and
long-term mortality. Furthermore, other hypotheses for the
larger long-term mortality with 80% oxygen also deserve atten-
tion, such as increased risk of cardiovascular disease. Hyperoxia
result in perioperative arterial vasoconstriction and reduced
coronary blood flow,33 34 and together with oxidative stress

i79
BJA Meyhoff et al.

Table 4 Time between surgery and events. Time is presented as the duration in days between index surgery and the first occurring outcome only
among patients meeting the respective outcomes

Outcome 80% oxygen group (n5678) 30% oxygen group (n5699)

n Median (IQR) n Median (IQR)
Time to cancer (days)
Overall 144 335 (146 658) 150 434 (214 797)
No cancer (n264 vs 260) 16 493 (239 1196) 18 607 (248 776)
Localized cancer (n208 vs 221) 57 390 (176 677) 54 567 (324 913)
Metastasized cancer (n206 vs 218) 71 281 (116 532) 78 318 (150 741)
Time to death (days)
Overall 208 392 (138 835) 182 416 (82 874)
No cancer (n264 vs 260) 26 338 (7947) 21 183 (43 654)
Localized cancer (n208 vs 221) 70 373 (112 899) 57 585 (82 995)
Metastasized cancer (n206 vs 218) 112 405 (164 745) 104 414 (99 798)
Time to cancer or death (days)
Overall 280 311 (115 660) 268 363 (96 776)
No cancer (n264 vs 260) 36 358 (68 992) 34 376 (50 760)
Localized cancer (n208 vs 221) 102 319 (134 706) 90 484 (152 874)
Metastasized cancer (n206 vs 218) 142 298 (116 573) 144 310 (88 605)

localized disease, but this did not seem to influence the

shorter survival in the 80% oxygen group.
100

Authors contributions
80
C.S.M.: study design, data collection, analysis and interpret-
% cancer-free survival

ation of data, drafting of manuscript, and critical revision of

60
40
30% oxygen, No cancer
30% oxygen, Localized cancer
20
30% oxygen, Metastases
80% oxygen, No cancer
80% oxygen, Localized cancer
0 80% oxygen, Metastases
0 1 2 3
Years since operation
Fig 2 Cancer-free survival in subgroups according to baseline
disease.
the manuscript. L.N.J.: study design, analysis and interpret-
ation of data, and critical revision of the manuscript. J.W.:
study design, analysis and interpretation of data, and critical
revision of the manuscript. V.D.S.: statistical analysis and critic-
al revision of the manuscript. L.S.R.: study design, analysis and
interpretation of data, and critical revision of the manuscript.
Declaration of interest
None declared.
4 5
Funding
This work was not directly supported, but the Tryg Foundation
provided funding (L.S.R.), and a number of grants were originally
received to conduct the PROXI trial.13

and DNA damage, this could lead to accelerated atheroscler-
osis or plaque rupture.2 3 35
We thus recommend avoiding administration of inspiratory
oxygen fractions higher than what is needed to maintain
normal arterial oxygen saturation until a benefit has been
documented.
In conclusion, no difference in new or recurrent cancers
was found in patients given 80% or 30% oxygen during and
2 h after abdominal surgery. Time to cancer may be shorter
after exposure to 80% oxygen, especially in patients with
References
1 Nunn JF. Oxygen. Applied Respiratory Physiology. London: Butter-
worth & Co Ltd, 1977; 375 444
2 Meyhoff CS, Staehr AK, Rasmussen LS. Rational use of oxygen in
medical disease and anesthesia. Curr Opin Anaesthesiol 2012; 25:
363 70
3 Lumb AB, Walton LJ. Perioperative oxygen toxicity. Anesthesiol Clin
2012; 30: 591 605
4 Martin DS, Grocott MP III. Oxygen therapy in anaesthesia: the yin
and yang of O2. Br J Anaesth 2013; 111: 867 71
5 Austin MA, Wills KE, Blizzard L, Walters EH, Wood-Baker R. Effect of
high flow oxygen on mortality in chronic obstructive pulmonary

i80
80% oxygen and risk of subsequent cancer
disease patients in prehospital setting: randomised controlled trial.
Br Med J 2010; 341: c5462
6 Kilgannon JH, Jones AE, Shapiro NI, et al. Association between ar-
terial hyperoxia following resuscitation from cardiac arrest and
in-hospital mortality. J Am Med Assoc 2010; 303: 216571
7 Brenner M, Stein D, Hu P, Kufera J, Wooford M, Scalea T. Association
between early hyperoxia and worse outcomes after traumatic brain
injury. Arch Surg 2012; 147: 10425
8 Cabello JB, Burls A, Emparanza JI, Bayliss S, Quinn T. Oxygen therapy
for acute myocardial infarction. Cochrane Database Syst Rev 2013;
8: CD007160
9 Hopf HW, Holm J. Hyperoxia and infection. Best Pract Res Clin Anaes-
thesiol 2008; 22: 553 69
10 Belda FJ, Aguilera L, Garca de la Asuncion J, et al. Supplemental
perioperative oxygen and the risk of surgical wound infection: a ran-
domized controlled trial. J Am Med Assoc 2005; 294: 203542
11 Greif R, Akca O, Horn EP, Kurz A, Sessler DI. Supplemental periopera-
tive oxygen to reduce the incidence of surgical-wound infection.
Outcomes Research Group. N Engl J Med 2000; 342: 161 7
12 Myles PS, Leslie K, Chan MT, et al. Avoidance of nitrous oxide for
patients undergoing major surgery: a randomized controlled trial.
Anesthesiology 2007; 107: 221 31
13 Meyhoff CS, Wetterslev J, Jorgensen LN, et al. Effect of high peri-
operative oxygen fraction on surgical site infection and pulmonary
complications after abdominal surgery: the PROXI randomized clin-
ical trial. J Am Med Assoc 2009; 302: 154350
14 Anthony T, Murray BW, Sum-Ping JT, et al. Evaluating an evidence-
based bundle for preventing surgical site infection: a randomized
trial. Arch Surg 2011; 146: 2639
15 Pryor KO, Fahey TJ III, Lien CA, Goldstein PA. Surgical site infection
and the routine use of perioperative hyperoxia in a general surgical
population: a randomized controlled trial. J Am Med Assoc 2004;
291: 79 87
16 Scifres CM, Leighton BL, Fogertey PJ, Macones GA, Stamilio DM. Sup-
plemental oxygen for the prevention of postcesarean infectious
morbidity: a randomized controlled trial. Am J Obstet Gynecol
2011; 205: 2679
17 Meyhoff CS, Jorgensen LN, Wetterslev J, Christensen KB,
Rasmussen LS. Increased long-term mortality after a high periopera-
tive inspiratory oxygen fraction during abdominal surgery: follow-up
of a randomized clinical trial. Anesth Analg 2012; 115: 84954
18 Hopf HW, Gibson JJ, Angeles AP, et al. Hyperoxia and angiogenesis.
Wound Repair Regen 2005; 13: 55864
19 De Bels D, Corazza F, Germonpre P, Balestra C. The normobaric
oxygen paradox: a novel way to administer oxygen as an adjuvant
treatment for cancer? Med Hypotheses 2011; 76: 46770
BJA
20 Meyhoff CS, Wetterslev J, Jorgensen LN, et al. Perioperative oxygen
fractioneffect on surgical site infection and pulmonary complica-
tions after abdominal surgery: a randomized clinical trial. Rationale
and design of the PROXI-Trial. Trials 2008; 9: 58
21 Andersen TF, Madsen M, Jorgensen J, Mellemkjoer L, Olsen JH. The
Danish National Hospital Register. A valuable source of data for
modern health sciences. Dan Med Bull 1999; 46: 2638
22 Kahan BC, Morris TP. Reporting and analysis of trials using stratified
randomisation in leading medical journals: review and reanalysis.
Br Med J 2012; 345: e5840
23 Sirop S, Kanaan M, Korant A, et al. Detection and prognostic impact
of micrometastasis in colorectal cancer. J Surg Oncol 2011; 103:
534 7
24 Sessler DI. Long-term consequences of anesthetic management.
Anesthesiology 2009; 111: 14
25 Benjamini Y, Hochberg Y. Controlling the false discovery rate: a prac-
tical and powerful approach to multiple testing. J R Stat Soc Ser
1995; 57: 289300
26 Qadan M, Battista C, Gardner SA, Anderson G, Akca O, Polk HC Jr.
Oxygen and surgical site infection: a study of underlying immuno-
logic mechanisms. Anesthesiology 2010; 113: 36977
27 Tertil M, Jozkowicz A, Dulak J. Oxidative stress in tumor angio-
genesistherapeutic targets. Curr Pharm Des 2010; 16: 3877 94
28 Raa A, Stansberg C, Steen VM, Bjerkvig R, Reed RK, Stuhr LE. Hyper-
oxia retards growth and induces apoptosis and loss of glands and
blood vessels in DMBA-induced rat mammary tumors. BMC
Cancer 2007; 7: 23
29 Bitterman H. Bench-to-bedside review: oxygen as a drug. Crit Care
2009; 13: 205
30 Leslie K, Myles PS, Chan MT, et al. Nitrous oxide and long-term mor-
bidity and mortality in the ENIGMA trial. Anesth Analg 2011; 112:
387 93
31 Gottschalk A, Ford JG, Regelin CC, et al. Association between epi-
dural analgesia and cancer recurrence after colorectal cancer
surgery. Anesthesiology 2010; 113: 27 34
32 Myles PS, Peyton P, Silbert B, Hunt J, Rigg JR, Sessler DI. Perioperative
epidural analgesia for major abdominal surgery for cancer and
recurrence-free survival: randomised trial. Br Med J 2011; 342:
d1491
33 Anderson KJ, Harten JM, Booth MG, et al. The cardiovascular effects
of normobaric hyperoxia in patients with heart rate fixed by per-
manent pacemaker. Anaesthesia 2010; 65: 167 71
34 Rothen HU. Oxygen: avoid too much of a good thing! Eur J Anaesthe-
siol 2010; 27: 4934
35 Glass CK, Witztum JL. Atherosclerosis. The road ahead. Cell 2001;
104: 503 16
Handling editor: D. J. Buggy
i81