Bioethics Paper

T h e n e w e n g l a n d j o u r n a l o f m e d i c i n e

n engl j med 376;23 nejm.org June 8, 2017 2223

The members of the writing committee
(Kathryn M. Rowan, Ph.D., Derek C. An-
gus, M.D., M.P.H., Michael Bailey, Ph.D.,
Amber E. Barnato, M.D., Rinaldo Bellomo,
M.D., Ruth R. Canter, M.Sc., Timothy J.
Coats, M.D., Anthony Delaney, M.D.,
Ph.D., Elizabeth Gimbel, R.N., B.S., Rich-
ard D. Grieve, Ph.D., David A. Harrison,
Ph.D., Alisa M. Higgins, M.P.H., Belinda
Howe, M.P.H., David T. Huang, M.D.,
M.P.H., John A. Kellum, M.D., Paul R.
Mouncey, M.Sc., Edvin Music, M.S.I.S.,
Sandra L. Peake, M.D., Ph.D., Francis
Pike, Ph.D., Michael C. Reade, M.B., B.S.,
D.Phil., M. Zia Sadique, Ph.D., Mervyn
Singer, M.D., and Donald M. Yealy, M.D.)
assume responsibility for the overall con-
tent and integrity of this article. The affili-
ations of the writing committee mem-
bers are listed in the Appendix. Address
reprint requests to Dr. Rowan at the In-
tensive Care National Audit and Research
Centre, Napier House, 24 High Holborn,
London WC1V 6AZ, United Kingdom, or
at kathy . rowan@ icnarc . org.

* The Protocolized Resuscitation in Sep-
sis Meta-Analysis (PRISM) study is a
collaboration of the Protocolized Care
for Early Septic Shock (ProCESS) Inves-
tigators, based in the United States; the
Australasian Resuscitation in Sepsis
Evaluation (ARISE) Investigators, based
in Australia and New Zealand; the Proto-
colised Management in Sepsis (ProMISe)
Investigators, based in the United King-
dom; and the International Forum for
Acute Care Trialists. A complete list of
the investigator groups is provided in
the Supplementary Appendix, available
at NEJM.org.

This article was published on March 21,
2017, at NEJM.org.

N Engl J Med 2017;376:2223-34.
DOI: 10.1056/NEJMoa1701380
Copyright © 2017 Massachusetts Medical Society.

BACKGROUND
After a single-center trial and observational studies suggesting that early, goal-
directed therapy (EGDT) reduced mortality from septic shock, three multicenter
trials (ProCESS, ARISE, and ProMISe) showed no benefit. This meta-analysis of
individual patient data from the three recent trials was designed prospectively to
improve statistical power and explore heterogeneity of treatment effect of EGDT.

METHODS
We harmonized entry criteria, intervention protocols, outcomes, resource-use
measures, and data collection across the trials and specified all analyses before
unblinding. After completion of the trials, we pooled data, excluding the protocol-
based standard-therapy group from the ProCESS trial, and resolved residual differ-
ences. The primary outcome was 90-day mortality. Secondary outcomes included
1-year survival, organ support, and hospitalization costs. We tested for treatment-
by-subgroup interactions for 16 patient characteristics and 6 care-delivery charac-
teristics.

RESULTS
We studied 3723 patients at 138 hospitals in seven countries. Mortality at 90 days
was similar for EGDT (462 of 1852 patients [24.9%]) and usual care (475 of 1871
patients [25.4%]); the adjusted odds ratio was 0.97 (95% confidence interval, 0.82
to 1.14; P = 0.68). EGDT was associated with greater mean (±SD) use of intensive
care (5.3±7.1 vs. 4.9±7.0 days, P = 0.04) and cardiovascular support (1.9±3.7 vs.
1.6±2.9 days, P = 0.01) than was usual care; other outcomes did not differ signifi-
cantly, although average costs were higher with EGDT. Subgroup analyses showed
no benefit from EGDT for patients with worse shock (higher serum lactate level,
combined hypotension and hyperlactatemia, or higher predicted risk of death) or
for hospitals with a lower propensity to use vasopressors or fluids during usual
resuscitation.

CONCLUSIONS
In this meta-analysis of individual patient data, EGDT did not result in better
outcomes than usual care and was associated with higher hospitalization costs
across a broad range of patient and hospital characteristics. (Funded by the Na-
tional Institute of General Medical Sciences and others; PRISM ClinicalTrials.gov
number, NCT02030158.)

A B S T R A C T

Early, Goal-Directed Therapy for Septic Shock
— A Patient-Level Meta-Analysis

The PRISM Investigators*

Original Article

The New England Journal of Medicine
Downloaded from nejm.org on April 22, 2021. For personal use only. No other uses without permission.

Copyright © 2017 Massachusetts Medical Society. All rights reserved.

n engl j med 376;23 nejm.org June 8, 20172224

T h e n e w e n g l a n d j o u r n a l o f m e d i c i n e

In 2001, Rivers and colleagues report-ed on a 263-patient, single-center, random-ized, controlled trial of early, goal-directed
therapy (EGDT) versus usual care in patients
presenting with septic shock to an urban emer-
gency department in the United States.1 EGDT is
a 6-hour resuscitation protocol for the adminis-
tration of intravenous fluids, vasopressors, ino-
tropes, and red-cell transfusion to achieve pre-
specified targets for arterial blood pressure,
central venous pressure, central venous oxygen
saturation, and hemoglobin level. EGDT reduced
hospital mortality from 46.5% to 30.5%,1 prompt-
ing many institutions worldwide to adopt EGDT.2
Three subsequent, government-funded, multi-
center, randomized, controlled trials from the
United States (Protocolized Care for Early Septic
Shock [ProCESS]),3 Australasia (Australasian Re-
suscitation in Sepsis Evaluation [ARISE]),4 and
the United Kingdom (Protocolised Management
in Sepsis [ProMISe])5 failed to show lower mor-
tality with EGDT than with usual care.

A meta-analysis combining the average re-
sults of the trials also indicated no overall bene-
fit from EGDT.6 There is considerable heteroge-
neity, however, in patients in whom septic shock
develops and in usual care across hospitals; con-
sequently, important treatment effects in patient
subgroups or particular settings may have been
missed.7

A prospective meta-analysis of individual pa-
tient data would provide greater statistical power
to identify subgroup effects. The ProCESS, ARISE,
and ProMISe investigators therefore planned this
prospective meta-analysis of individual patient
data (called the Protocolized Resuscitation in
Sepsis Meta-Analysis [PRISM] study) before en-
rollment of the first patient into the first trial
and harmonized entry criteria, intervention pro-
tocols, outcomes, major resource-use measures,
and data collection across the three trials.8 The
goals of the current study were to use pooled
data from the three trials to determine the effect
of EGDT versus usual care on 90-day mortality
and secondary clinical and economic outcomes
and to compare the effects of EGDT across pre-
specified patient and care-delivery subgroups.

M e t h o d s

Study Design

All three trials evaluated the EGDT protocol, as
described in the article by Rivers et al.1 Core

aspects of best care, including early recognition
of sepsis and prompt delivery of intravenous
fluids and antimicrobial agents, were promoted
in the EGDT groups and the usual-care groups
and reinforced through trial eligibility criteria.

We published the statistical analysis plan and
a priori hypotheses for the current study before
unblinding of any results from the three trials
(ClinicalTrials.gov number, NCT02030158); the
protocol is also available with the full text of
this article at NEJM.org. Each trial supplied in-
dividual patient data after publication3-5 and after
the trial-level meta-analysis.6 Before pooling data,
we compared trial protocols, case-report forms,
and data dictionaries to identify any recoding
needed. We then provided a detailed data-set
specification to each trial team to prepare the
data file for pooling. After receipt of the data,
we checked for missing or duplicate values and
for consistency and plausibility, resolving data
queries through direct consultation with each
trial team before analysis. We did not reassess
risk of bias because that had been performed for
the trial-level meta-analysis.6

The final data-set specification is shown in
Table S1 in the Supplementary Appendix, avail-
able at NEJM.org. The primary outcome measure
was all-cause mortality at 90 days. Secondary
outcome measures were in-hospital and 28-day
mortality; duration of survival to 1 year; dura-
tion of stay in the emergency department, inten-
sive care unit, and hospital; receipt and duration
of invasive mechanical ventilation, vasopressors,
and renal-replacement therapy; and costs and
cost-effectiveness at 90 days.

Prespecified subgroups according to baseline
patient characteristics were age, sex, severe co-
existing conditions (liver, respiratory, cardiovas-
cular, and renal conditions and immunocom-
promised state, all defined according to Acute
Physiology and Chronic Health Evaluation
[APACHE] II criteria), site of infection, and sever-
ity of illness. Severity of illness was operational-
ized in eight ways, according to eligibility crite-
ria met (refractory hypotension, hyperlactatemia,
or both), serum lactate level, illness-severity score
(APACHE II Acute Physiology Score [range, 0 to
60, with higher scores indicating greater severity
of illness] and APACHE II score [range, 0 to 71,
with higher scores indicating greater severity of
illness]), organ dysfunction (Sequential Organ
Failure Assessment score), treatment (invasive
mechanical ventilation [yes or no] and vasopres-

The New England Journal of Medicine
Downloaded from nejm.org on April 22, 2021. For personal use only. No other uses without permission.

Copyright © 2017 Massachusetts Medical Society. All rights reserved.

n engl j med 376;23 nejm.org June 8, 2017 2225

E a r ly, G oa l-Dir ec ted Ther a py for Sep tic Shock

sors [yes or no]), and risk of death (derived from
a customized model; see the Supplementary Ap-
pendix). Prespecified subgroups according to
care-delivery characteristics were time from
emergency department presentation to random-
ization, time of randomization (weekday or week-
end and day or night), time from emergency
department presentation to first administration
of intravenous antimicrobial agents (available for
the ProCESS and ARISE trials), and underlying
intensity of care (derived from propensity mod-
els for the use of vasopressors or fluids during
usual care; see the Supplementary Appendix).

The funders had no role in the design or con-
duct of the study, in the collection, analysis, or
interpretation of the data, or in the writing of
the manuscript or the decision to submit it for
publication.

Statistical Analysis

The individual trials each had 80 to 90% power
to detect an absolute difference in mortality of
6.5 to 8.0 percentage points between the EGDT
group and the usual-care group, under the as-
sumption of a baseline mortality of 24 to 40%,
depending on the trial. Because this was a pro-
spective meta-analysis of individual patient data,
the sample-size calculation was undertaken be-
fore the results of the individual trials were
available. On the basis of a control event rate of
25 to 35%, a statistical power of 80%, and a
two-sided P value of 0.05 (with no allowance for
heterogeneity of treatment effect or clustering of
outcomes across trials), this study could detect
an absolute between-group difference in 90-day
mortality of 4 to 5 percentage points and an
interaction effect (odds ratio) of approximately
1.5 or 1.6 for a subgroup representing one half
or one quarter of the total sample, respectively.

We conducted all analyses on an intention-to-
treat basis. We used one-stage, hierarchical re-
gression modeling (patients nested in sites nested
in trials), with site as a random effect and trial
as a fixed effect. We determined heterogeneity
among trials by fitting a fixed interaction be-
tween treatment and trial. We analyzed binomial
outcomes using hierarchical logistic regression,
reported as odds ratios and 95% confidence in-
tervals; survival time (censored at 1 year) using
hierarchical (shared frailty) Cox proportional-
hazards regression, reported as hazard ratios
and 95% confidence intervals; and continuous
outcomes using hierarchical linear regression,

reported as differences in means and 95% con-
fidence intervals. We presented survival to 1 year
using a Kaplan–Meier survival curve.

We performed a secondary analysis of the
primary outcome using the same hierarchical
regression structure with adjustment for pre-
specified baseline covariates of age, sex, last
systolic blood pressure before randomization
(<90 or ≥90 mm Hg), APACHE II score, and in- vasive mechanical ventilation at randomization (yes or no). Analyses of binomial secondary out- comes were adjusted for the same covariates. To determine heterogeneity between prespecified subgroups, we added fixed interaction terms between treatment and subgroup to the adjusted model for the primary outcome. To ascertain whether any variation in treatment effect across subgroups was consistent among the trials, we fitted three-way fixed interactions among trial, treatment, and subgroup. We analyzed continu- ous subgroup variables by dividing the cohort into thirds. Our cost-effectiveness analysis compared the outcomes and costs, from the health-services perspective, up to 90 days after randomization. We used the combined mortality but reported cost and cost-effectiveness estimates separately for each trial because the interpretation of pooled cost-effectiveness estimates is unclear when drawn from health care systems with different cost structures.9 The resource use for each pa- tient was combined with trial-specific unit costs to report the incremental costs of EGDT versus usual care. We calculated quality-adjusted life- years (QALYs) up to 90 days by combining sur- vival time with quality-of-life scores from the EuroQol questionnaire (EQ-5D-5L) administered at 90 days in the ProMISe trial, using the area- under-the-curve approach.10 We estimated incre- mental costs and QALYs of EGDT versus usual care with a seemingly unrelated regression model,11 with trial as a fixed effect for costs. We report results for each trial overall and for the same prespecified subgroups as for the clinical outcomes. We report incremental net monetary benefits by valuing QALYs at recommended thresholds for a QALY gain and performed sen- sitivity analyses to test the robustness of our results to alternative assumptions (see the Sup- plementary Appendix). All analyses were performed with the use of SAS software, version 9.4 (SAS Institute), or Stata software, version 11.2 (StataCorp), and a two- The New England Journal of Medicine Downloaded from nejm.org on April 22, 2021. For personal use only. No other uses without permission. Copyright © 2017 Massachusetts Medical Society. All rights reserved. n engl j med 376;23 nejm.org June 8, 20172226 T h e n e w e n g l a n d j o u r n a l o f m e d i c i n e Characteristic EGDT (N = 1857) Usual Care (N = 1880) Patient characteristics Age — yr† Median 65 65 IQR 53–75 53–76 Male sex — no. (%) 1065 (57.4) 1104 (58.7) ≥1 Severe coexisting condition — no./total no. (%)‡ 546/1854 (29.4) 526/1880 (28.0) Site of infection — no. (%) Lungs 657 (35.4) 620 (33.0) Abdomen 172 (9.3) 163 (8.7) Blood 172 (9.3) 172 (9.1) Central nervous system 28 (1.5) 19 (1.0) Soft tissue 154 (8.3) 153 (8.1) Urinary tract 356 (19.2) 371 (19.7) Other 113 (6.1) 149 (7.9) Unknown 196 (10.6) 218 (11.6) Determined ultimately to have no infection 9 (0.5) 15 (0.8) Entry criterion met — no./total no. (%) Refractory hypotension only 821/1854 (44.3) 833/1880 (44.3) Hyperlactatemia only 717/1854 (38.7) 732/1880 (38.9) Both refractory hypotension and hyperlactatemia 316/1854 (17.0) 315/1880 (16.8) Last values before randomization Systolic blood pressure — mm Hg Median 94 94 IQR 83–112 82–111 Mean arterial pressure — mm Hg Median 67 67 IQR 59–78 59–78 Serum lactate — mmol/liter Median 4.3 4.2 IQR 2.5–5.9 2.4–5.9 APACHE II Acute Physiology Score — median (IQR)§ 11 (7–15) 11 (7–15) APACHE II score — median (IQR)¶ 16 (12–21) 16 (12–21) SOFA score — median (IQR)‖ 4 (2–6) 4 (2–6) Customized risk of death — median (IQR) 0.21 (0.11–0.37) 0.22 (0.11–0.36) Care-delivery characteristics Time from ED presentation to inclusion criteria met — min Median 85 81 IQR 40–150 36–145 Time from ED presentation to randomization — min Median 162 159 IQR 119–223 115–221 Receiving antimicrobial agents at randomization — no./total no. (%) 1726/1856 (93.0) 1742/1880 (92.7) Table 1. Patient and Care-Delivery Characteristics at Baseline.* The New England Journal of Medicine Downloaded from nejm.org on April 22, 2021. For personal use only. No other uses without permission. Copyright © 2017 Massachusetts Medical Society. All rights reserved. n engl j med 376;23 nejm.org June 8, 2017 2227 E a r ly, G oa l-Dir ec ted Ther a py for Sep tic Shock sided alpha level of 0.05. Complete-case analysis was used for clinical outcomes because data were missing for less than 0.5% for all out- comes; multiple imputation was used for miss- ing quality-of-life scores. We did not adjust for multiple comparisons; with 22 planned subgroup analyses, 1 or 2 significant interaction tests (P<0.05) would be expected on the basis of chance alone.12 R e s u l t s Study Patients From March 2008 through July 2014, the three trials enrolled 4211 patients at 138 hospitals in the United States (ProCESS); Australia, New Zea- land, Finland, Hong Kong, and the Republic of Ireland (ARISE); and England (ProMISe). The 448 patients randomly assigned to receive proto- col-based standard therapy in the ProCESS trial were excluded from the current study, resulting in 3763 patients randomly assigned to either usual care (1892 patients) or EGDT (1871 patients). After the exclusion of patients who withdrew consent, underwent randomization in error, or were lost to follow-up at 90 days, 3723 patients (98.9%) were included in the primary analysis and 3511 (93.3%) were followed up to 1 year (Fig. S1 in the Supplementary Appendix). Patient and care-delivery characteristics were well bal- anced at baseline (Table 1, and Tables S2 and S3 in the Supplementary Appendix). Primary Outcome Mortality at 90 days did not differ significantly between the two groups. Death occurred in 462 of 1852 patients (24.9%) in the EGDT group and in 475 of 1871 (25.4%) in the usual-care group (Table 2). The adjusted odds ratio was 0.97 (95% confidence interval [CI], 0.82 to 1.14; P = 0.68). Characteristic EGDT (N = 1857) Usual Care (N = 1880) Time from ED presentation to first IV antimicrobial agents — min** Median 75 72 IQR 42–120 42–119 IV fluids administered before hospital presentation until randomization — no./total no. (%) 1801/1846 (97.6) 1818/1871 (97.2) Volume administered — ml Median 2000 2000 IQR 1250–3000 1200–3000 Volume administered per kilogram of body weight — ml Median 27.5 27.7 IQR 16.5–42.3 16.2–41.7 * Data are from the Protocolized Care for Early Septic Shock (ProCESS) trial, the Australasian Resuscitation in Sepsis Evaluation (ARISE) trial, and the Protocolised Management in Sepsis (ProMISe) trial. The numbers of patients with data available for analysis were as follows: age, 1857 in the group that received early, goal-directed therapy (EGDT) and 1879 in the group that received usual care; systolic blood pressure, 1809 and 1824; mean arterial pressure, 1318 and 1352; serum lactate, 1626 and 1645; customized risk of death, 1849 and 1878; time from emergency department (ED) presentation to inclusion criterion met, 1853 and 1878; time from ED presentation to first intravenous (IV) anti- microbial agents, 1091 and 1095; volume of IV fluids administered, 1846 and 1871; and volume of IV fluids adminis- tered per kilogram of body weight, 1723 and 1687. For details on data harmonization, see Table S1 in the Supplementary Appendix. IQR denotes interquartile range. † Age was estimated for 7 patients in the ProMISe trial. ‡ Severe coexisting conditions were defined according to Acute Physiology and Chronic Health Evaluation [APACHE] II criteria. § APACHE II Acute Physiology Scores range from 0 to 60, with higher scores indicating greater severity of illness. ¶ APACHE II scores range from 0 to 71, with higher scores indicating greater severity of illness. ‖ Scores on the Sequential Organ Failure Assessment (SOFA) range from 0 to 24, with higher scores indicating a greater degree of organ failure. Baseline urine output was not used in the calculation of the renal SOFA score in the ARISE and ProMISe trials. ** Shown are data for patients who received IV antimicrobial agents before randomization in the ProCESS and ARISE trials. All patients in the ProMISe trial received IV antimicrobial agents before randomization (time not recorded). Table 1. (Continued.) The New England Journal of Medicine Downloaded from nejm.org on April 22, 2021. For personal use only. No other uses without permission. Copyright © 2017 Massachusetts Medical Society. All rights reserved. n engl j med 376;23 nejm.org June 8, 20172228 T h e n e w e n g l a n d j o u r n a l o f m e d i c i n e T ab le 2 . O u tc o m es .* O u tc o m e E G D T ( N = 1 85 7) U su al C ar e (N = 1 88 0) In cr em en ta l E ff ec t (9 5% C I) P V al u e O ve ra ll C o m p ar is o n C o m p ar is o n am o n g T ri al s P ri m ar y o u tc o m e: d ea th a t 90 d ay s — n o ./ to ta l n o . ( % ) 46 2/ 18 52 ( 24 .9 ) 47 5/ 18 71 ( 25 .4 ) 0. 97 ( 0. 82 t o 1 .1 4) † ‡ 0. 68 0. 73 S ec o n d ar y o u tc o m es : m o rt al it y D ea th a t h o sp it al d is ch ar ge — n o ./ to ta l n o . ( % )§ 37 0/ 18 57 ( 19 .9 ) 36 5/ 18 78 ( 19 .4 ) 1. 02 ( 0. 85 t o 1 .2 1) † 0. 86 0. 42 D ea th a t 28 d ay s — n o ./ to ta l n o . ( % ) 37 5/ 18 54 ( 20 .2 ) 38 5/ 18 73 ( 20 .6 ) 0. 96 ( 0. 81 t o 1 .1 5) † 0. 68 0. 57 S ec o n d ar y o u tc o m es : d u ra ti o n o f st ay f ro m r an d o m iz at io n In E D — h r M ed ia n 1 1 IQ R 0 to 3 0 to 3 M ea n 2. 1± 3. 3 2. 2± 3. 0 −0 .1 ( −0 .3 t o 0 .1 )¶ 0. 19 < 0. 00 1‖ In I C U A d m it te d t o I C U — n o . ( % ) 16 84 ( 90 .7 ) 15 32 ( 81 .5 ) F ir st s ta y — d ay s M ed ia n a m o n g p at ie n ts a d m it te d 3 4 IQ R 2 to 6 2 to 6 M ea n o ve ra ll 4. 9± 6. 6 4. 5± 6. 4 0. 5 (0 .1 t o 0 .9 )¶ 0. 02 0. 76 T o ta l s ta y, in cl u d in g re ad m is si o n s — d ay s M ed ia n a m o n g p at ie n ts a d m it te d 4 4 IQ R 2 to 7 2 to 7 M ea n o ve ra ll 5. 3± 7. 1 4. 9± 7. 0 0. 5 (0 .0 t o 0 .9 )¶ 0. 04 0. 78 In h o sp it al — d ay s§ M ed ia n 9 9 IQ R 5 to 1 7 5 to 1 7 M ea n 14 .8 ± 17 .5 14 .9 ± 26 .2 −0 .1 ( −1 .5 t o 1 .4 )¶ 0. 92 0. 39 S ec o n d ar y o u tc o m es : r ec ei p t an d d u ra ti o n o f o rg an s u p p o rt in I C U R es p ir at o ry s u p p o rt : i n va si ve m ec h an ic al v en ti la ti o n in I C U R ec ei p t — n o ./ to ta l n o . ( % ) 56 5/ 18 52 ( 30 .5 ) 54 4/ 18 74 ( 29 .0 ) 1. 05 ( 0. 89 t o 1 .2 4) † 0. 57 0. 04 ** D u ra ti o n — d ay s M ed ia n a m o n g p at ie n ts r ec ei vi n g su p p o rt 4 4 IQ R 2 to 8 2 to 8 M ea n o ve ra ll 2. 1± 5. 5 1. 9± 5. 2 0. 2 (− 0. 2 to 0 .5 )¶ 0. 36 0. 58 The New England Journal of Medicine Downloaded from nejm.org on April 22, 2021. For personal use only. No other uses without permission. Copyright © 2017 Massachusetts Medical Society. All rights reserved. n engl j med 376;23 nejm.org June 8, 2017 2229 E a r ly, G oa l-Dir ec ted Ther a py for Sep tic Shock O u tc o m e E G D T ( N = 1 85 7) U su al C ar e (N = 1 88 0) In cr em en ta l E ff ec t (9 5% C I) P V al u e O ve ra ll C o m p ar is o n C o m p ar is o n am o n g T ri al s C ar d io va sc u la r su p p o rt : v as o p re ss o rs o r in o tr o p es in I C U R ec ei p t — n o ./ to ta l n o . ( % ) 10 40 /1 85 4 (5 6. 1) 92 3/ 18 73 ( 49 .3 ) 1. 42 ( 1. 23 t o 1 .6 4) † < 0. 00 1 0. 40 D u ra ti o n — d ay s M ed ia n a m o n g p at ie n ts r ec ei vi n g su p p o rt 2 2 IQ R 1 to 4 1 to 4 M ea n o ve ra ll 1. 9± 3. 7 1. 6± 2. 9 0. 3 (0 .1 t o 0 .5 )¶ 0. 01 0. 52 R en al s u p p o rt : r en al -r ep la ce m en t th er ap y in I C U R ec ei p t — n o ./ to ta l n o . ( % ) 20 4/ 18 52 ( 11 .0 ) 19 8/ 18 74 ( 10 .6 ) 1. 02 ( 0. 81 t o 1 .2 8) † 0. 88 0. 91 D u ra ti o n — d ay s M ed ia n a m o n g p at ie n ts r ec ei vi n g su p p o rt 3 4 IQ R 2 to 7 2 to 7 M ea n o ve ra ll 0. 7± 3. 3 0. 6± 2. 4 0. 0 (− 0. 1 to 0 .2 )¶ 0. 68 0. 99 C o st -e ff ec ti ve n es s an al ys is † † T o ta l c o st s u p t o 9 0 d ay s — $ P ro C E S S 32 ,1 78 ± 30 ,1 81 30 ,9 30 ± 30 ,1 50 12 76 ( −1 79 9 to 4 35 2) ¶ 0. 42 A R IS E 25 ,0 14 ± 25 ,7 37 22 ,9 73 ± 22 ,8 22 20 42 ( −2 64 t o 4 35 2) ¶ 0. 08 P ro M IS e 14 ,1 12 ± 15 ,1 20 12 ,9 06 ± 16 ,0 17 11 83 ( −1 41 8 to 3 78 3) ¶ 0. 37 E Q -5 D -5 L sc o re a m o n g su rv iv o rs a t 90 d ay s‡ ‡ 0. 62 3± 0. 31 3 0. 62 5± 0. 30 9 −0 .0 02 ( −0 .0 39 t o 0 .0 00 )¶ 0. 91 Q A LY s am o n g al l p at ie n ts t o 9 0 d ay s 0. 05 8± 0. 04 8 0. 05 8± 0. 04 8 0. 00 0 (− 0. 00 4 to 0 .0 04 )¶ 0. 96 In cr em en ta l n et b en ef it a t 90 d ay s — $ §§ P ro C E S S −1 26 6 (− 43 73 t o 1 84 1) 0. 43 A R IS E −2 03 2 (− 43 78 t o 3 14 ) 0. 09 P ro M IS e −1 17 2 (− 38 13 t o 1 46 9) 0. 39 * P lu s– m in u s va lu es a re m ea n s ± S D . F o r d et ai ls o n d at a h ar m o n iz at io n , se e T ab le S 1 in t h e S u p p le m en ta ry A p p en d ix . C I d en o te s co n fi d en ce i n te rv al , IC U i n te n si ve c ar e u n it , an d Q A LY q u al it y- ad ju st ed l if e- ye ar . † S h o w n i s th e ad ju st ed o d d s ra ti o , w it h a d ju st m en t fo r ag e, s ex , la st s ys to lic b lo o d p re ss u re b ef o re r an d o m iz at io n ( < 90 o r ≥ 90 m m H g) , A P A C H E I I sc o re , an d r ec ei p t o f in va si ve m e- ch an ic al v en ti la ti o n ( ye s o r n o ). ‡ T h e u n ad ju st ed o d d s ra ti o f o r th e p ri m ar y o u tc o m e w as 0 .9 8 (9 5% C I, 0 .8 4 to 1 .1 4; P = 0 .7 8) . § D at a w er e ce n so re d a t 60 d ay s af te r ra n d o m iz at io n i n t h e P ro C E S S t ri al . ¶ S h o w n i s th e d if fe re n ce i n m ea n s. ‖ T h e in cr em en ta l e ff ec t (d if fe re n ce in m ea n s) a cc o rd in g to t ri al w as a s fo llo w s: P ro C E S S , 0 .3 ( 95 % C I, − 0. 1 to 0 .7 ); A R IS E , − 0. 6 (9 5% C I, − 0. 9 to − 0. 3) ; a n d P ro M IS e, 0 .2 ( 95 % C I, 0 .0 t o 0 .4 ). ** T h e in cr em en ta l ef fe ct ( ad ju st ed o d d s ra ti o ) ac co rd in g to t ri al w as a s fo llo w s: P ro C E S S , 1. 51 ( 95 % C I, 1 .1 2 to 2 .0 4) ; A R IS E , 0. 98 ( 95 % C I, 0 .7 8 to 1 .2 3) ; an d P ro M IS e, 0 .9 8 (9 5% C I, 0. 76 t o 1 .2 6) . † † M is si n g d at a w er e m u lt ip ly i m p u te d . ‡ ‡ Q u al it y o f lif e w as a ss es se d w it h t h e u se o f th e E u ro Q o l q u es ti o n n ai re ( E Q -5 D -5 L; a s co re o f 0 in d ic at es d ea th a n d 1 p er fe ct q u al it y o f lif e) , w h ic h w as a d m in is te re d t o e lig ib le p at ie n ts in t h e P ro M IS e tr ia l at 9 0 d ay s af te r ra n d o m iz at io n . F o r al l p …