Estudio TRISS

The n e w e ng l a n d j o u r na l of m e dic i n e
original article
Lower versus Higher Hemoglobin Threshold

for Transfusion in Septic Shock
Lars B. Holst, M.D., Nicolai Haase, M.D., Ph.D., Jørn Wetterslev, M.D., Ph.D.,
Jan Wernerman, M.D., Ph.D., Anne B. Guttormsen, M.D., Ph.D.,
Sari Karlsson, M.D., Ph.D., Pär I. Johansson, M.D., Ph.D.,
Anders Åneman, M.D., Ph.D., Marianne L. Vang, M.D., Robert Winding, M.D.,
Lars Nebrich, M.D., Helle L. Nibro, M.D., Ph.D., Bodil S. Rasmussen, M.D., Ph.D.,
Johnny R.M. Lauridsen, M.D., Jane S. Nielsen, M.D., Anders Oldner, M.D., Ph.D.,
Ville Pettilä, M.D., Ph.D., Maria B. Cronhjort, M.D., Lasse H. Andersen, M.D.,
Ulf G. Pedersen, M.D., Nanna Reiter, M.D., Jørgen Wiis, M.D.,
Jonathan O. White, M.D., Lene Russell, M.D., Klaus J. Thornberg, M.D.,
Peter B. Hjortrup, M.D., Rasmus G. Müller, M.D., Morten H. Møller, M.D., Ph.D.,
Morten Steensen, M.D., Inga Tjäder, M.D., Ph.D., Kristina Kilsand, R.N.,
Suzanne Odeberg-Wernerman, M.D., Ph.D., Brit Sjøbø, R.N.,
Helle Bundgaard, M.D., Ph.D., Maria A. Thyø, M.D., David Lodahl, M.D.,
Rikke Mærkedahl, M.D., Carsten Albeck, M.D., Dorte Illum, M.D., Mary Kruse, M.D.,
Per Winkel, M.D., D.M.Sci., and Anders Perner, M.D., Ph.D.,
for the TRISS Trial Group* and the Scandinavian Critical Care Trials Group
From the Department of Intensive Care
A BS T R AC T (L.B.H., N.H., L.H.A., U.G.P., N.R., J. Wiis,
J.O.W., L.R., K.J.T., P.B.H., R.G.M., M.H.M.,
BACKGROUND M.S., A.P.), Copenhagen Trial Unit, Center
for Clinical Intervention Research (J. Wet-
Blood transfusions are frequently given to patients with septic shock. However, the terslev, P.W.), and Section for Transfu-
benefits and harms of different hemoglobin thresholds for transfusion have not sion Medicine (P.I.J.), Rigshospitalet and
been established. University of Copenhagen, Copenhagen,
Randers Hospital, Randers (M.L.V., H.B.,
METHODS M.A.T.), Herning Hospital, Herning (R.W.,
In this multicenter, parallel-group trial, we randomly assigned patients in the inten- D.L., R.M.), Hvidovre Hospital, Hvidovre
(L.N., C.A.), Aarhus University Hospital,
sive care unit (ICU) who had septic shock and a hemoglobin concentration of 9 g Aarhus (H.L.N., D.I.), Aalborg University
per deciliter or less to receive 1 unit of leukoreduced red cells when the hemoglobin Hospital, Aalborg (B.S.R.), Holbæk Hospi-
level was 7 g per deciliter or less (lower threshold) or when the level was 9 g per tal, Holbæk (J.R.M.L.), Kolding Hospital,
deciliter or less (higher threshold) during the ICU stay. The primary outcome mea- Kolding (J.S.N.), and Hjørring Hospital,
Hjørring (M.K.) — all in Denmark; Karo-
sure was death by 90 days after randomization. linska University Hospital, Huddinge, Stock
RESULTS holm (J. Wernerman, I.T., K.K., S.O.-W.),
Karolinska University Hospital, Solna (A.O.),
We analyzed data from 998 of 1005 patients (99.3%) who underwent randomiza- and Södersjukhuset, Stockholm (M.B.C.)
tion. The two intervention groups had similar baseline characteristics. In the ICU, — all in Sweden; Haukeland University
the lower-threshold group received a median of 1 unit of blood (interquartile range, Hospital and University of Bergen, Bergen,
Norway (A.B.G., B.S.); Tampere Univer-
0 to 3) and the higher-threshold group received a median of 4 units (interquartile sity Hospital, Tampere (S.K.), and Hel-
range, 2 to 7). At 90 days after randomization, 216 of 502 patients (43.0%) assigned sinki University Hospital and University of
to the lower-threshold group, as compared with 223 of 496 (45.0%) assigned to the Helsinki, Helsinki (V.P.) — all in Finland;
higher-threshold group, had died (relative risk, 0.94; 95% confidence interval, 0.78 and Liverpool Hospital, Sydney (A.Å.).
Address reprint requests to Dr. Perner at
to 1.09; P = 0.44). The results were similar in analyses adjusted for risk factors at the Department of Intensive Care, Rigshos-
baseline and in analyses of the per-protocol populations. The numbers of patients pitalet, Blegdamsvej 9, DK-2100 Copen-
who had ischemic events, who had severe adverse reactions, and who required life hagen, Denmark, or at anders.perner@
regionh.dk.
support were similar in the two intervention groups.
* Members of the Transfusion Require-
CONCLUSIONS ments in Septic Shock (TRISS) Trial
Among patients with septic shock, mortality at 90 days and rates of ischemic events Group are listed in the Supplementary
Appendix, available at NEJM.org.
and use of life support were similar among those assigned to blood transfusion at a
This article was published on October 1,
higher hemoglobin threshold and those assigned to blood transfusion at a lower
2014, at NEJM.org.
threshold; the latter group received fewer transfusions. (Funded by the Danish Stra- DOI: 10.1056/NEJMoa1406617
tegic Research Council and others; TRISS ClinicalTrials.gov number, NCT01485315.) Copyright © 2014 Massachusetts Medical Society.
n engl j med nejm.org 1

The New England Journal of Medicine
Downloaded from nejm.org at PURDUE UNIVERSITY LIB TSS on October 1, 2014. For personal use only. No other uses without permission.
Copyright © 2014 Massachusetts Medical Society. All rights reserved.
B
lood transfusions are frequently sent was withdrawn or not granted, we asked the
given to patients with septic shock.1-4 Some patient or surrogate for permission to continue
of these transfusions are given to patients registration of trial data and to use these data in
who are bleeding, but many nonbleeding patients the analyses. The protocol, including details re-
also undergo transfusion.5 garding trial conduct and the statistical analysis
The recommendations of the Surviving Sepsis plan, has been published previously15 and is avail-
Campaign regarding blood transfusion in pa- able with the full text of this article at NEJM.org.
tients with septic shock are complex and include The management committee (see the Supplemen-
a recommendation for transfusion to maintain a tary Appendix, available at NEJM.org) designed
hematocrit of more than 30% in the presence of the trial and vouches for the adherence of the
hypoperfusion in the first 6 hours.6 After that, study to the protocol and for the accuracy of the
the transfusion threshold should be a hemoglo- data and the analyses. The members of the man-
bin level of less than 7 g per deciliter, aiming at agement committee wrote the drafts of the manu-
levels between 7 g and 9 g per deciliter in pa- script and made the decision to submit the man-
tients who do not have myocardial ischemia, uscript for publication. The funders had no role
severe hypoxemia, acute hemorrhage, or ischemic in the design of the protocol, the trial conduct, or
coronary artery disease.6 However, there are the analyses or reporting of the data.
limited data supporting these recommendations,6 This trial was a multicenter, stratified, parallel-
and many clinicians may not follow them.4,7 group, clinical trial. Randomization was per-
New trial data have been published recently,8 formed with the use of a centralized computer-
and the use of a high hemoglobin threshold for generated assignment sequence, with stratification
transfusion may be at least questioned as part of according to study site and the presence or ab-
an early resuscitation protocol for patients with sence of active hematologic cancer, because these
septic shock. characteristics may influence outcome.16,17 Pa-
Blood transfusion has been associated with tients with septic shock were randomly assigned
increased mortality in subgroups of critically ill in a 1:1 ratio, with the use of permuted blocks
patients, both in cohort studies and in random- of varying sizes of 6, 8, or 10, to blood transfu-
ized trials,9-12 but there have also been cohort sion at the higher hemoglobin threshold or the
studies in which transfusion was associated with lower hemoglobin threshold. Treatment assign-
improved survival,13 including among patients ments were concealed from the investigators
with sepsis.14 In some studies, nonleukoreduced assessing mortality, the data and safety monitor-
blood was used, which may have influenced the ing committee, and the trial statistician. The
results. Given the lack of efficacy data, in addition conduct of the trial and the safety of the par-
to concerns about safety, we conducted the Trans- ticipants were overseen by the data and safety
fusion Requirements in Septic Shock (TRISS) monitoring committee, which performed an in-
trial to evaluate the effects on mortality of leuko- terim analysis after 500 patients had been fol-
reduced blood transfusion at a lower versus a lowed for 90 days. The trial data were monitored
higher hemoglobin threshold among patients by staff from the coordinating center.
with septic shock who are in the intensive care
unit (ICU). TRIAL PATIENTS
We screened patients 18 years of age or older who
ME THODS were in the ICU, fulfilled the criteria for septic
shock,18 and had a blood concentration of hemo-
TRIAL DESIGN AND OVERSIGHT globin of 9 g per deciliter or less as measured by
After the approvals from ethics committees and means of valid point-of-care testing (see the Sup-
data-protection agencies were obtained, patients in plementary Appendix). The reasons for the exclu-
32 general ICUs in Denmark, Sweden, Norway, and sion of some patients are shown in Figure 1 and
Finland underwent screening and randomization listed in the Supplementary Appendix.
between December 3, 2011, and December 26,
2013. Written informed consent was obtained INTERVENTION
from all the patients or their legal surrogates be- Enrolled patients were given single units of cross-
fore or after enrollment. In all cases, consent was matched, prestorage leukoreduced red cells sus-
obtained from the patient when possible. If con- pended in a saline–adenine–glucose–mannitol
2 n engl j med nejm.org

Hemoglobin Threshold for Tr ansfusion in Septic Shock
1224 Patients were assessed for eligibility
219 Were excluded

3 Declined transfusion
4 Had previous adverse reaction
to transfusion
137 Received blood transfusion in ICU
20 Had acute coronary syndrome
16 Had life-threatening bleeding
3 Had acute burn injury
17 Withdrew from active therapy
34 Were excluded because consent could
not be obtained
1005 Underwent randomization
5 Were excluded after randomization

1 Underwent randomization in error
4 Withdrew consent
503 Were assigned to the lower 497 Were assigned to the higher
hemoglobin threshold hemoglobin threshold
38 Discontinued the study 24 Discontinued the study

29 Were withdrawn at 18 Were withdrawn at
patient’s or surrogate’s patient’s or surrogate’s
request request
9 Were withdrawn at 6 Were withdrawn at
physician’s request physician’s request
1 Withdrew consent for the 1 Withdrew consent for the
use of data use of data
502 (99.8%) Were included in all analyses 496 (99.8%) Were included in all analyses
of mortality of mortality
488 (97.0%) Were included in all analyses 489 (98.4%) Were included in all analyses
of outcomes of outcomes
Figure 1. Assessment, Randomization, and Follow-up.

Patients were excluded if they had undergone randomization in this study previously, if there were medical reasons,
if they had received a blood transfusion during the current intensive care unit (ICU) admission, if there was a docu-
mented wish not to receive a transfusion, or if informed consent could not be obtained. A total of 15 patients met
two exclusion criteria. One patient was excluded immediately after randomization when it was determined that an
inclusion criterion had not been met, and 4 were excluded because consent was withdrawn during the trial. Thereafter,
5 additional patients underwent randomization in order for the study to obtain the full sample. All the patients who with-
drew from the trial at their own request or at a surrogate’s request allowed the use of their data, but 14 patients or
surrogates in the lower-threshold group (hemoglobin level, ≤7 g per deciliter) and 7 in the higher-threshold group
(hemoglobin level, ≤9 g per deciliter) did not want further data registered except for mortality data, which were obtained
from national registries. The process data (hemoglobin assessments and numbers of transfusions and temporary
protocol suspensions and protocol violations) and some of the secondary-outcome data for these patients are missing.
solution when the blood concentration of hemo- shock.15 Hemoglobin concentrations were reas-
globin had decreased to the assigned transfusion sessed within 3 hours after termination of the
threshold (≤7 g per deciliter [lower threshold] or transfusion or before the initiation of another
≤9 g per deciliter [higher threshold]). These lev- transfusion. The intervention period was the en-
els of hemoglobin have frequently been used as tire ICU stay, to a maximum of 90 days after ran-
thresholds for transfusion in patients with septic domization.

In the event that life-threatening bleeding or of 5%, assuming a mortality in the higher-
ischemia developed while a patient was in the threshold group of 45% (estimated from two pre-
ICU or a patient required the use of extracorpo- vious cohorts).20,21 The estimated difference of
real membrane oxygenation, the patient could 9 percentage points was derived from the 20%
receive a transfusion at a hemoglobin threshold reduction in relative risk observed with a restric-
decided by the attending doctor. The attending tive versus liberal transfusion strategy in the sub-
doctor decided when the patient again was to group of patients with severe infection in the
receive a transfusion at the assigned hemoglobin Transfusion Requirements in Critical Care
threshold. After the unmasking of trial data (TRICC) trial.9 During our trial, 5 patients were
showing harm from hydroxyethyl starch,3 we excluded after randomization (4 patients did not
recommended against the use of all starch prod- allow the use of their data, and 1 did not have
ucts in trial patients. All other interventions were sepsis, which was realized immediately after ran-
at the discretion of the clinicians, including trans- domization). A total of 5 additional patients un-
fusion during surgery and after ICU discharge. derwent randomization in order for the study to
obtain the full sample (Fig. 1).
OUTCOME MEASURES An author who was the statistician for the
The primary outcome measure was death by 90 study and who was unaware of the study-group
days after randomization. Secondary outcome assignments performed all the analyses accord-
measures were the use of life support (defined as ing to International Conference on Harmonisa-
the use of vasopressor or inotropic therapy, me- tion Good Clinical Practice guidelines22 and the
chanical ventilation, or renal-replacement ther- statistical analysis plan.15 We performed the
apy) at days 5, 14, and 28 after randomization19; primary analyses in the intention-to-treat popu-
the number of patients with serious adverse reac- lation, which included all the patients who under-
tions while in the ICU (allergic reaction, hemoly- went randomization, except for those whose data
sis, transfusion-associated acute lung injury, or were deleted from the database during the trial
transfusion-associated circulatory overload) (see (i.e., the 5 patients, noted above, who were ex-
the Supplementary Appendix); the number of pa- cluded after randomization) and after the trial
tients with ischemic events while in the ICU, (2 patients who withdrew consent for the use of
which included cerebral ischemia (identified from their data) (Fig. 1). In the per-protocol popula-
the results of imaging), acute myocardial ische tions, we excluded patients who had one or more
mia (defined by symptoms, electrocardiographic bleeding or ischemic episodes or one or more
signs, or elevated biomarker levels resulting in an major protocol violations (see the Supplementary
intervention), intestinal ischemia (as observed dur- Appendix).22
ing endoscopic examination or surgery), or limb In the primary analyses (including the analy-
ischemia (defined as clinical signs resulting in an sis of the primary outcome measure), we com-
intervention) (for full definitions, see the Supple- pared data between the two groups by means of
mentary Appendix); the percentage of days alive logistic-regression analysis for binary outcome
without vasopressor or inotropic therapy, mechan- measures with adjustment for the stratification
ical ventilation, or renal-replacement therapy in variables (study site and presence or absence of
the 90 days after randomization; and the per- active hematologic cancer),23 and we converted
centage of days alive and out of the hospital in odds ratios to relative risks.24 We also performed
the 90 days after randomization. Data for the unadjusted chi-square testing for binary outcome
outcome measures were obtained by TRISS trial measures and Wilcoxon signed-rank testing for
investigators or their delegates from patient files rate and ordinal data. We compared the primary
and national and regional registries for the entire outcome in the per-protocol populations and in
90-day follow-up period. prespecified subgroups defined according to the
presence or absence of chronic cardiovascular
STATISTICAL ANALYSIS disease (i.e., any history of myocardial infarc-
We calculated that we would need to enroll 1000 tion, any history of stable or unstable angina
patients for the trial to have 80% power to show pectoris, previous treatment with nitrates, percu-
mortality at 90 days that was 9 percentage points taneous coronary intervention, coronary-artery
lower in the lower-threshold group than in the bypass grafting or noncoronary vascular interven-
higher-threshold group, at a two-sided alpha level tions, any history of chronic heart failure [defined

as New York Heart Association class III or IV], lower-threshold group did not undergo transfu-
or any history of cerebral infarction or transitory sion in the ICU, as compared with 6 (1.2%) in the
cerebral ischemia), an age of 70 years or younger higher-threshold group (P<0.001). Details regard-
versus an age older than 70 years, and a Simpli- ing blood products, bleeding, cointerventions,
fied Acute Physiology Score (SAPS) II above 53 fluid volumes and balances, and circulatory as-
versus 53 or lower at baseline (with the score sessments are provided in Tables S4 through S9
calculated from 17 variables and ranging from in the Supplementary Appendix. The numbers of
0 to 163, with higher scores indicating higher protocol violations differed significantly be-
severity of disease) and used multiple logistic- tween the two groups (Table S10 in the Supple-
regression analyses in the intention-to-treat pop- mentary Appendix).
ulation to adjust for differences in prespecified
risk factors at baseline. Details regarding the OUTCOMES
handling of missing data are provided in the At 90 days after randomization, 216 patients
Supplementary Appendix. We performed all (43.0%) in the lower-threshold group and 223
analyses using SAS software, version 9.3 (SAS (45.0%) in the higher-threshold group had died
Software), and SPSS software, version 17.0 (SPSS). (relative risk, 0.94; 95% confidence interval, 0.78
A two-sided P value of less than 0.05 was consid- to 1.09; P = 0.44) (Table 2 and Fig. 3, and Table
ered to indicate statistical significance. S11 in the Supplementary Appendix). We ob-
tained similar results in the analyses that were
R E SULT S adjusted for prespecified baseline risk factors
and in the per-protocol analyses (Table S12 in the
TRIAL POPULATION Supplementary Appendix). The prespecified sub-
We obtained 90-day vital status for 998 patients group analyses showed no significant heteroge-
(99.3%), including 502 in the lower-threshold neity in the effect of the transfusion threshold on
group and 496 in the higher-threshold group mortality at 90 days between patients with and
(Fig. 1). The characteristics of the patients at those without chronic cardiovascular disease,
baseline were similar in the two groups (Table 1, patients 70 years of age or younger and those
and Table S1 in the Supplementary Appendix). A older than 70 years of age, and patients with a
total of 29 of 488 patients (5.9%) in the lower- SAPS II of 53 or less and those with a SAPS II of
threshold group and 11 of 489 (2.2%) in the more than 53 at baseline (Fig. 3).
higher-threshold group had the protocol tempo- A total of 7.2% of the patients in the lower-
rarily suspended (P = 0.004) (Table S2 in the Sup- threshold group, as compared with 8.0% in the
plementary Appendix). higher-threshold group, had one or more ische
mic events in the ICU (Table 2, and Tables S13
HEMOGLOBIN CONCENTRATIONS, BLOOD and S14 in the Supplementary Appendix, which
PRODUCTS, AND CIRCULATORY VARIABLES include the numbers of patients with myocardial
The median value of the lowest concentration of ischemia and ischemia of other anatomical sites).
hemoglobin in the 24 hours before randomiza- One patient had a serious adverse reaction to
tion was 8.4 g per deciliter in both intervention transfusion (Table 2, and Table S13 in the Sup-
groups. After randomization, the daily lowest con- plementary Appendix). The use of life support at
centrations of hemoglobin differed between the days 5, 14, and 28 was similar in the two inter-
two groups (P<0.001) (Fig. 2). Additional details vention groups (Table 2, and Tables S11 and S13
regarding hemoglobin assessments are provided in the Supplementary Appendix), as were the
in Table S3 in the Supplementary Appendix. percentages of days alive without vasopressor or
During the trial period, a total of 1545 blood inotropic therapy, without mechanical ventila-
transfusions were given in the lower-threshold tion, and without renal-replacement therapy and
group and 3088 transfusions in the higher- the percentage of days alive and out of the hos-
threshold group (P<0.001). The median cumula- pital (Table 2).
tive number of blood transfusions after random-
ization was 1 unit (interquartile range, 0 to 3) in DISCUSSION
the lower-threshold group and 4 (interquartile
range, 2 to 7) in the higher-threshold group In this international, multicenter, partially blind-
(P<0.001). A total of 176 patients (36.1%) in the ed, randomized trial involving patients with sep-

tic shock who were in the ICU, we observed no ceived 50% fewer units of blood than those in the
significant differences in mortality at 90 days, in higher-threshold group, and 36% of the patients
the numbers of patients with ischemic events or in the lower-threshold group did not undergo
with severe adverse reactions, in the use of life transfusion in the ICU, as compared with 1% of
support, or in the numbers of days alive and out the patients in the higher-threshold group.
of the hospital between the group of patients Our results are consistent with those ob-
who underwent transfusion at a lower hemoglo- tained in the TRICC trial, which assessed a
bin threshold and the group of those who under- lower versus higher hemoglobin threshold for
went transfusion at a higher hemoglobin thresh- blood transfusion in a broad population of adult
old. Similar results were observed in subgroups patients in the ICU.9 In that trial, there were no
of patients with chronic cardiovascular disease, significant differences in mortality at 30 days in
with older age, or with greater disease severity. the full trial population (the primary outcome)
The patients in the lower-threshold group re- or among patients 55 years of age or older or
Table 1. Characteristics of the Trial Patients at Baseline.*
Lower Hemoglobin Higher Hemoglobin

Threshold Threshold
Characteristic (N = 502) (N = 496)

Age — yr
Median 67 67
Interquartile range 57–73 58–75
Male sex — no. (%) 272 (54.2) 259 (52.2)
Chronic cardiovascular disease — no. (%)† 75 (14.9) 66 (13.3)
Chronic lung disease — no. (%)‡ 111 (22.1) 102 (20.6)
Hematologic cancer — no. (%) 39 (7.8) 36 (7.3)
Admission to a university hospital — no. (%) 323 (64.3) 324 (65.3)
Surgery during index hospitalization — no. (%)
Emergency 191 (38.0) 217 (43.8)
Elective 59 (11.8) 53 (10.7)
Source of ICU admittance — no. (%)
Emergency department 90 (17.9) 79 (15.9)
General ward 268 (53.4) 257 (51.8)
Operating or recovery room 113 (22.5) 121 (24.4)
Other ICU 31 (6.2) 39 (7.9)
Source of sepsis — no. (%)§
Lungs 267 (53.2) 259 (52.2)
Abdomen 206 (41.0) 198 (39.9)
Urinary tract 58 (11.6) 61 (12.3)
Soft tissue 59 (11.8) 59 (11.9)
Other 50 (10.0) 47 (9.5)
Positive culture from blood or sterile site 188 (37.5) 160 (32.3)
Interval from ICU admission to randomization — hr
Median 23 20
SAPS II¶
Median 51 52

Table 1. (Continued.)
Lower Hemoglobin Higher Hemoglobin

Threshold Threshold
Characteristic (N = 502) (N = 496)

SOFA score‖
Median 10 10
Renal-replacement therapy — no. (%)** 68 (13.5) 53 (10.7)
Mechanical ventilation — no. (%)†† 345 (68.7) 350 (70.6)
* None of the differences between the two groups were significant (P≥0.05). Additional details regarding baseline char-
acteristics are provided in Table S1 in the Supplementary Appendix. The lower hemoglobin threshold was defined as
a hemoglobin level of 7 g per deciliter or less, and the higher hemoglobin threshold as a hemoglobin level of 9 g per
deciliter or less. ICU denotes intensive care unit.
† Patients were considered to have chronic cardiovascular disease if they had any history of myocardial infarction, sta-
ble or unstable angina pectoris, chronic heart failure (defined as New York Heart Association class III or IV), cerebral
infarction or transitory cerebral ischemia, previous treatment with nitrates, percutaneous coronary intervention, coro-
nary-artery bypass grafting, or noncoronary vascular interventions.
‡ Patients were considered to have chronic lung disease if they had any history of chronic obstructive pulmonary dis-
ease, asthma or other chronic lung disease, or any treatment with a drug indicated for chronic lung disease.
§ Some patients had more than one source of infection. Other sources of sepsis included a vascular catheter, meningi-
tis, or endocarditis or were unclear.
¶ The Simplified Acute Physiology Score (SAPS) II25 was assessed in the 24 hours before randomization. The SAPS II is
calculated from 17 variables and ranges from 0 to 163, with higher scores indicating higher severity of disease. One
or two of the 17 variables were missing for 77 patients in the higher-threshold group and for 99 in the lower-threshold
group, so their values were not included here.
‖ The Sepsis-Related Organ Failure Assessment (SOFA)26 score was assessed in the 24 hours before randomization.
The SOFA grades organ failure, with subscores ranging from 0 to 4 for each of six organ systems (cerebral, circula-
tion, pulmonary, hepatic, renal, and coagulation). The aggregated score ranges from 0 to 24, with higher scores indi-
cating more severe organ failure. One variable was missing for 51 patients in the higher-threshold group and for 64 in
the lower-threshold group, so their values were not included here.
** Renal-replacement therapy was defined as therapy for acute or chronic kidney failure at randomization.
†† Mechanical ventilation was defined as invasive or noninvasive ventilation in the 24 hours before randomization.
11
Higher hemoglobin threshold

10
Blood Hemoglobin (g/dl)
7
Lower hemoglobin threshold
0
Base- 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
line
Days since Randomization
Figure 2. Blood Hemoglobin Levels in Patients in the ICU at Baseline and after Randomization.
The graphs show the median daily lowest levels of blood hemoglobin in the lower-threshold group and the higher-
threshold group. Baseline values were the lowest blood hemoglobin level measured in the 24 hours before random-
ization. Day 1 was defined as the time of randomization to the end of that day and lasted a median of 15 hours in
the lower-threshold group and 14 hours in the higher-threshold group. The I bars indicate the 25th and 75th percentiles.

Table 2. Primary and Secondary Outcome Measures.*
Lower Hemoglobin Higher Hemoglobin Relative Risk

Outcome Threshold Threshold (95% CI) P Value
Primary outcome: death by day 90 — no./total no. (%) 216/502 (43.0) 223/496 (45.0) 0.94 (0.78–1.09) 0.44†
Secondary outcomes‡
Use of life support — no./total no. (%)§
At day 5 278/432 (64.4) 267/429 (62.2) 1.04 (0.93–1.14) 0.47†
At day 14 140/380 (36.8) 135/367 (36.8) 0.99 (0.81–1.19) 0.95†
At day 28 53/330 (16.1) 64/322 (19.9) 0.77 (0.54–1.09) 0.14†
Ischemic event in the ICU — no./total no. (%)¶ 35/488 (7.2) 39/489 (8.0) 0.90 (0.58–1.39) 0.64‖
Severe adverse reaction — no./total no. (%)** 0/488 1/489 (0.2) — 1.00
Alive without vasopressor or inotropic therapy — 73 75 — 0.93
mean % of days††
Alive without mechanical ventilation — mean % 65 67 — 0.49
of days††
Alive without renal-replacement therapy — mean % 85 83 — 0.54
of days††
Alive and out of the hospital — mean % of days†† 30 31 — 0.89
* CI denotes confidence interval.

† Logistic-regression analyses were adjusted for the stratification variables (study site and presence or absence of hematologic cancer). The
results of the unadjusted outcome analyses are provided in Table S11 in the Supplementary Appendix.
‡ A total of 21 patients — 14 in the lower-threshold group and 7 in the higher-threshold group — did not wish to be included in the follow-
up, so data regarding secondary outcome measures are missing for these patients.
§ Use of life support was defined as infusion of vasopressor or inotropic agents or the use of invasive or noninvasive mechanical ventilation
or renal-replacement therapy on those days. The total number of patients decreased because patients died. See Table S13 in the Supple
mentary Appendix.
¶ An ischemic event in the ICU was defined as one or more events of acute myocardial, cerebral, intestinal, or limb ischemia. See Table S13
in the Supplementary Appendix.
‖ Logistic-regression analyses were adjusted for the presence of hematologic cancer. Adjustment according to study site was not possible,
because there were zero events at four study sites.
** A severe adverse reaction was defined as allergic reaction, hemolysis, transfusion-associated acute lung injury, or transfusion-associated
circulatory overload. See Table S13 in the Supplementary Appendix.
†† The mean percentage of days was calculated as the number of days without vasopressor, ventilator, or renal-replacement therapy, divided
by the number of days alive during the 90-day follow-up period, or as the number of days out of the hospital, divided by the number of
days alive during the 90-day follow-up period.
those with more severe disease; these two sub- the three trials. In the TRICC trial, significantly
groups may best resemble our patients. Our re- increased rates of myocardial infarction were
sults are also in line with those of a large trial observed with a higher transfusion threshold,9
involving high-risk patients after hip surgery, whereas the opposite was observed in the FOCUS
the Transfusion Trigger Trial for Functional Out- trial and in our trial, although the numerical
comes in Cardiovascular Patients Undergoing differences were not significant in either of these
Surgical Hip Fracture Repair (FOCUS) trial,27 two trials.27 In our trial, myocardial infarction
and the Cochrane meta-analysis of trials of trans- was not a prespecified outcome measure (the
fusion thresholds, both of which support restric- data are provided in the Supplementary Appen-
tive transfusion to reduce the use of blood in dix); we did not specify surveillance testing for
patients with preexisting cardiovascular disease.28 myocardial ischemia in the protocol and may
An important exception is patients with acute myo- have missed some events. This may also have
cardial infarction, who were excluded both from resulted in detection bias because the clinicians
our trial and from the FOCUS trial.27 Research and investigators were not unaware of the inter-
is needed to assess the safety of lower hemoglo- vention assignments.
bin thresholds for transfusion in these patients.12 We observed no harm with an excess transfu-
The effect of transfusion thresholds on rates sion of a median of 3 units of blood, a finding
of myocardial infarction may have differed among that is contrary to most of the observational data

A Time to Death
1.0
0.8
Probability of Survival
Lower hemoglobin threshold
0.6
Higher hemoglobin threshold
0.4
0.2
P=0.41
0.0
0 10 20 30 40 50 60 70 80 90
Days since Randomization
No. at Risk
Lower hemoglobin threshold 502 334 306 286
Higher hemoglobin threshold 496 321 287 273
B Relative Risk of the Primary Outcome

Lower Higher P Value
Hemoglobin Hemoglobin for Hetero-
Subgroup Threshold Threshold Relative Risk (95% CI) geneity
no. of events/no. of patients in subgroup
Age 0.85
>70 yr 93/173 98/185 0.98 (0.79–1.18)
≤70 yr 123/329 125/311 0.94 (0.75–1.14)
Chronic cardiovascular 0.25
disease
Yes 42/75 33/66 1.08 (0.75–1.40)
No 174/427 190/430 0.90 (0.75–1.06)
SAPS II at baseline 0.06
>53 112/207 139/226 0.83 (0.64–1.04)
≤53 104/295 84/270 1.10 (0.91–1.30)
All patients 216/502 223/496 0.94 (0.78–1.09)
0.5 0.7 1.0 1.5 2.0
Lower Higher
Hemoglobin Hemoglobin
Threshold Threshold
Better Better
Figure 3. Time to Death and Relative Risk of Death at 90 Days.

Panel A shows the survival curves, with data censored at 90 days, in the two intervention groups in the intention-to-
treat population. Kaplan–Meier analysis showed that the survival time did not differ significantly between the two groups
(P = 0.41 by Cox regression analysis, with adjustment for the stratification variables). Panel B shows the relative risks
(black boxes) with 95% confidence intervals (horizontal lines) for the primary outcome measure of death by day 90 in
the lower-threshold group, as compared with the higher-threshold group, among all the patients and in the three pre-
specified subgroups, as assessed by means of logistic-regression analysis, with adjustment for the stratification vari-
ables. The size of each black box is proportional to the size of the corresponding subgroup. Chronic cardiovascular
disease was defined as any history of myocardial infarction, any history of stable or unstable angina pectoris, previous
treatment with nitrates, percutaneous coronary intervention, coronary-artery bypass grafting or noncoronary vascular
interventions, any history of chronic heart failure (defined as New York Heart Association class III or IV), or any history
of cerebral infarction or transitory cerebral ischemia. The Simplified Acute Physiology Score (SAPS) II25 is calculated
from 17 baseline variables; scores range from 0 to 163, with higher scores indicating higher severity of disease. A total
of 1 or 2 of the 17 variables were missing for 77 patients in the higher-threshold group and for 99 in the lower-threshold
group. In this analysis, these missing variables were considered to be within the normal range, thereby not contributing
to the composite SAPS II of these patients.

regarding transfusion in critically ill patients.10 differences in some other outcome measures (in
Whether this was due to the use of leukoreduced particular, the ischemic events) and in some of
blood cannot be assessed, but results similar to the subgroup analyses (in particular, the sub-
ours were observed in the FOCUS trial, in which group defined according to the presence or ab-
the majority of patients also received leukore- sence of chronic cardiovascular disease).
duced blood.27 The safety of leukoreduced blood We recorded only one serious adverse reaction
was challenged by the results of a trial involving to blood transfusion, but serious adverse reac-
patients with upper gastrointestinal bleeding, tions are rare events in general, and their fre-
which showed increased mortality with liberal quencies are unknown among patients with sep-
transfusion of this product.11 Ongoing bleeding tic shock in the ICU. We included some patients
may have contributed to the increased mortality who had received a blood transfusion before ICU
observed with liberal transfusion in that trial.11 admission, and some patients had protocol sus-
Thus the effects of leukoreduction on outcome pensions and violations, which tended to reduce
are unclear, as they were a decade ago, as indi- the difference between the two intervention
cated in a 2004 meta-analysis of trial data on groups. However, we found clear differences be-
leukoreduced versus nonleukoreduced blood.29 tween the two groups in the hemoglobin levels
The strengths of our trial include a low risk and the numbers of transfusions, and the per-
of bias, because group assignment at random- protocol analyses, which excluded patients who
ization was concealed, and the blinding of the had protocol suspensions and violations, support
assessors of mortality and the statistician to the ed the primary analysis. Protocol suspensions
assigned intervention. It is reasonable to assume and violations have been difficult to prevent in
that our results are generalizable, because pa- transfusion trials,32,33 and when reported they
tients were recruited both in university hospitals appear to have occurred at frequencies similar to
and in nonuniversity hospitals, and the majority those observed in our trial.
of patients who underwent screening were includ In conclusion, patients with septic shock who
ed. The trial protocol was pragmatic, so routine underwent transfusion at a hemoglobin thresh-
practice was maintained except for the hemoglo- old of 7 g per deciliter, as compared with those
bin thresholds for transfusion. In addition, the who underwent transfusion at a hemoglobin
characteristics of the patients and the outcome threshold of 9 g per deciliter, received fewer
rates were similar to those observed in some transfusions and had similar mortality at 90 days,
recent trials involving patients with septic shock use of life support, and number of days alive and
in the ICU.3,19,30,31 out of the hospital; the numbers of patients with
Our trial has limitations. First, the investiga- ischemic events and severe adverse reactions to
tors, clinicians, and patients were aware of the blood in the ICU were also similar in the two
study-group assignments, and we did not assess intervention groups.
all the cointerventions. Because the trial was mul-
ticenter and large and used stratified random- Supported by a grant (09-066938) from the Danish Strategic
Research Council and by Copenhagen University Hospital, Rigs
ization, it is unlikely that imbalance in concomi- hospitalet, the Scandinavian Society of Anaesthesiology and In-
tant interventions affected the results. Second, tensive Care Medicine (ACTA Foundation), and Ehrenreich’s
the confidence interval was relatively wide for Foundation.
Dr. Johansson reports receiving grant support from Pharma-
the point estimate for mortality, so we cannot cosmos; and Dr. Perner, receiving grant support from CSL Beh-
exclude a 9% relative increase or a 22% relative ring, Fresenius Kabi, Cosmed, and Bioporto Diagnostics, and
decrease in mortality at 90 days in the lower- lecture fees from LFB. No other potential conflict of interest
relevant to this article was reported.
threshold group versus the higher-threshold Disclosure forms provided by the authors are available with
group. Third, we had limited power to detect the full text of this article at NEJM.org.
REFERENCES
1. Zarychanski R, Doucette S, Fergusson among patients with septic shock receiv- Med 2012;367:124-34. [Erratum, N Engl J
D, et al. Early intravenous unfractionated ing appropriate initial antibiotic treatment. Med 2012;367:481.]
heparin and mortality in septic shock. Crit Care Med 2012;40:2016-21. 4. Rosland RG, Hagen MU, Haase N,
Crit Care Med 2008;36:2973-9. 3. Perner A, Haase N, Guttormsen AB, et al. Red blood cell transfusion in septic
2. Labelle A, Juang P, Reichley R, et al. et al. Hydroxyethyl starch 130/0.4 versus shock — clinical characteristics and out-
The determinants of hospital mortality Ringer’s acetate in severe sepsis. N Engl J come of unselected patients in a prospec-

tive, multicentre cohort. Scand J Trauma vere sepsis and septic shock: a propensity- blocks or minimisation. Stat Med 2012;31:
Resusc Emerg Med 2014;22:14. matched analysis. Crit Care Med 2012;40: 328-40.
5. Haase N, Wetterslev J, Winkel P, Perner 3140-5. 24. Zhang J, Yu KF. What’s the relative
A. Bleeding and risk of death with hy- 15. Holst LB, Haase N, Wetterslev J, et al. risk? A method of correcting the odds ratio
droxyethyl starch in severe sepsis: post Transfusion requirements in septic shock in cohort studies of common outcomes.
hoc analyses of a randomized clinical (TRISS) trial — comparing the effects JAMA 1998;280:1690-1.
trial. Intensive Care Med 2013;39:2126-34. and safety of liberal versus restrictive red 25. Le Gall JR, Lemeshow S, Saulnier F.
6. Dellinger RP, Levy MM, Rhodes A, blood cell transfusion in septic shock — A new Simplified Acute Physiology Score
et al. Surviving Sepsis Campaign: interna- patients in the ICU: protocol for a ran (SAPS II) based on a European/North
tional guidelines for management of se- domised controlled trial. Trials 2013;14: American multicenter study. JAMA 1993;
vere sepsis and septic shock, 2012. Inten- 150. 270:2957-63. [Erratum, JAMA 1994;271:
sive Care Med 2013;39:165-228. 16. Vincent JL, Sakr Y, Sprung CL, et al. 1321.]
7. Reade MC, Huang DT, Bell D, et al. Sepsis in European intensive care units: 26. Vincent JL, Moreno R, Takala J, et al.
Variability in management of early severe results of the SOAP study. Crit Care Med The SOFA (sepsis-related organ failure
sepsis. Emerg Med J 2010;27:110-5. 2006;34:344-53. assessment) score to describe organ dys-
8. The ProCESS Investigators. A random- 17. Reinikainen M, Karlsson S, Varpula T, function/failure. Intensive Care Med 1996;
ized trial of protocol-based care for early et al. Are small hospitals with small in- 22:707-10.
septic shock. N Engl J Med 2014;370:1683- tensive care units able to treat patients 27. Carson JL, Terrin ML, Noveck H, et al.
93. with severe sepsis? Intensive Care Med Liberal or restrictive transfusion in high-
9. Hébert PC, Wells G, Blajchman MA, 2010;36:673-9. risk patients after hip surgery. N Engl J
et al. A multicenter, randomized, con- 18. American College of Chest Physicians/ Med 2011;365:2453-62.
trolled clinical trial of transfusion require- Society of Critical Care Medicine Consen- 28. Carson JL, Carless PA, Hebert PC.
ments in critical care. N Engl J Med 1999; sus Conference: definitions for sepsis and Transfusion thresholds and other strate-
340:409-17. [Erratum, N Engl J Med 1999; organ failure and guidelines for the use of gies for guiding allogeneic red blood cell
340:1056.] innovative therapies in sepsis. Crit Care transfusion. Cochrane Database Syst Rev
10. Marik PE, Corwin HL. Efficacy of red Med 1992;20:864-74. 2012;4:CD002042.
blood cell transfusion in the critically ill: 19. Heyland DK, Muscedere J, Drover J, 29. Fergusson D, Khanna MP, Tinmouth A,
a systematic review of the literature. Crit Jiang X, Day AG, Canadian Critical Care Hébert PC. Transfusion of leukoreduced
Care Med 2008;36:2667-74. [Erratum, Crit Trials Group. Persistent organ dysfunction red blood cells may decrease postopera-
Care Med 2008;36:3134.] plus death: a novel, composite outcome tive infections: two meta-analyses of ran-
11. Villanueva C, Colomo A, Bosch A, et al. measure for critical care trials. Crit Care domized controlled trials. Can J Anaesth
Transfusion strategies for acute upper 2011;15:R98. 2004;51:417-24.
gastrointestinal bleeding. N Engl J Med 20. Carlsen S, Perner A, East Danish Septic 30. Asfar P, Meziani F, Hamel J-F, et al.
2013;368:11-21. [Erratum, N Engl J Med Shock Cohort Investigators. Initial fluid High versus low blood-pressure target in
2013;368:2341.] resuscitation of patients with septic shock patients with septic shock. N Engl J Med
12. Chatterjee S, Wetterslev J, Sharma A, in the intensive care unit. Acta Anaesthe- 2014;370:1583-93.
Lichstein E, Mukherjee D. Association of siol Scand 2011;55:394-400. 31. Caironi P, Tognoni G, Masson S, et al.
blood transfusion with increased mortal- 21. Smith SH, Perner A. Higher vs. lower Albumin replacement in patients with se-
ity in myocardial infarction: a meta-analy- fluid volume for septic shock: clinical vere sepsis or septic shock. N Engl J Med
sis and diversity-adjusted study sequential characteristics and outcome in unselected 2014;370:1412-21.
analysis. JAMA Intern Med 2013;173:132-9. patients in a prospective, multicenter co- 32. Walsh TS, Boyd JA, Watson D, et al.
13. Vincent JL, Sakr Y, Sprung C, Harboe hort. Crit Care 2012;16:R76. Restrictive versus liberal transfusion strat-
S, Damas P, Sepsis Occurrence in Acutely 22. ICH Steering Committee. International egies for older mechanically ventilated
Ill Patients (SOAP) Investigators. Are blood conference on harmonisation of technical critically ill patients: a randomized pilot
transfusions associated with greater mor- requirements for registration of pharma- trial. Crit Care Med 2013;41:2354-63.
tality rates? Results of the Sepsis Occur- ceuticals for human use. ICH Harmon- 33. Rohde JM, Dimcheff DE, Blumberg N,
rence in Acutely Ill Patients study. Anes- ised Tripartite Guideline for Statistical et al. Health care-associated infection after
thesiology 2008;108:31-9. Principles for Clinical Trials. 3rd ed. Rich- red blood cell transfusion: a systematic
14. Park DW, Chun BC, Kwon SS, et al. mond, UK: Brookwood Medical, 1998. review and meta-analysis. JAMA 2014;311:
Red blood cell transfusions are associated 23. Kahan BC, Morris TP. Improper anal- 1317-26.
with lower mortality in patients with se- ysis of trials randomised using stratified Copyright © 2014 Massachusetts Medical Society.


Estudio TRISS

Uploaded by

Copyright:

Available Formats

Estudio TRISS

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Estudio TRISS

Uploaded by

Copyright:

Available Formats

The n e w e ng l a n d j o u r na l of m e dic i n e

Lower versus Higher Hemoglobin Threshold

n engl j med nejm.org 1

2 n engl j med nejm.org

The New England Journal of Medicine

1224 Patients were assessed for eligibility

219 Were excluded

1005 Underwent randomization

5 Were excluded after randomization

38 Discontinued the study 24 Discontinued the study

Figure 1. Assessment, Randomization, and Follow-up.

n engl j med nejm.org 3

4 n engl j med nejm.org

The New England Journal of Medicine

n engl j med nejm.org 5

Table 1. Characteristics of the Trial Patients at Baseline.*

Lower Hemoglobin Higher Hemoglobin

6 n engl j med nejm.org

The New England Journal of Medicine

Lower Hemoglobin Higher Hemoglobin

Higher hemoglobin threshold

n engl j med nejm.org 7

Table 2. Primary and Secondary Outcome Measures.*

Lower Hemoglobin Higher Hemoglobin Relative Risk

* CI denotes confidence interval.

8 n engl j med nejm.org

The New England Journal of Medicine

Higher hemoglobin threshold

B Relative Risk of the Primary Outcome

Figure 3. Time to Death and Relative Risk of Death at 90 Days.

n engl j med nejm.org 9

10 n engl j med nejm.org

The New England Journal of Medicine

n engl j med nejm.org 11

You might also like