Septic shock is associated with significant morbidity and mortality (1–3). Randomized clinical trials in critical care and sepsis have typically assessed short-term mortality, with fewer trials assessing long-term intervention effects on mortality and patient-centered outcomes such as health-related quality of life (HR-QoL). Although mortality has decreased over the past decade (4), recognition of long-term morbidity after survival from sepsis has shifted the focus from short-term mortality to longer-term patient-centered outcomes. A systematic review of HR-QoL in sepsis found that survivors consistently demonstrated impaired HR-QoL and recommended studies include longer-term endpoints (5).
The Australasian Resuscitation in Sepsis Evaluation (ARISE) trial was a multicenter, randomized controlled trial (RCT) to compare early goal-directed therapy (EGDT) with usual care (UC) in patients presenting to the emergency department (ED) with early septic shock. The trial found no difference in the primary outcome of all-cause mortality at 90 days postrandomization (6). This article presents the long-term mortality and HR-QoL for patients enrolled in the ARISE trial, as prespecified in the statistical analysis plan of the trial (7).
MATERIALS AND METHODS
Study Design and Participants
ARISE was a prospective, randomized, parallel-group trial conducted in 51 hospitals in Australia, New Zealand, Finland, Hong Kong, and the Republic of Ireland. The ARISE RCT design, methodology, and main results have been previously published (6). Briefly, patients presenting to the ED with early septic shock (defined as suspected or confirmed infection, two or more criteria for a systemic inflammatory response , and evidence of refractory hypotension or hypoperfusion) were randomized in a 1:1 ratio to receive EGDT or UC for a 6-hour period. Participants were randomized between October 2008 and April 2014.
Ethics approval was obtained at all participating hospitals and written informed consent was obtained from the patient or their legal representative. Ethics approval for patient follow-up to conduct HR-QoL assessments was obtained after 62 participants had been enrolled. Following completion of the trial, where site resources allowed, an application was made to ethics committees at participating sites to allow determination of survival status at 1-year following randomization in participants who were randomized prior to ethics approval for long-term follow-up or in participants who had declined consent to be contacted for HR-QoL assessments. At sites where this occurred and was approved, hospital records were searched to determine participants’ survival status at 1-year postrandomization.
For participants from whom consent was obtained, survival status at each time point was determined using medical records, or contact with the participant, their next of kin, or general practitioner. Where a participant was deceased, the date of death was recorded. HR-QoL was assessed using the Short Form 36 (SF-36) version 2, the EuroQoL-5D-3L quality of life (QoL) assessment tool (EQ-5D-3L) and the Assessment of Quality of Life 4D (AQoL). The EQ-5D-3L and SF-36 are recommended for measuring HR-QoL in critical care trials (9).
The EQ-5D-3L is a preference-based QoL instrument comprised of five dimensions: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. Respondents are asked to choose the most appropriate option from three alternatives (no, moderate, or severe problems). In addition, respondents are asked to indicate their present health state on a EuroQoL Visual Analog Scale (EQ VAS) ranging from the worst imaginable health state (“0”) to the best imaginable health state (“100”) (10). The EQ-5D-3L is a valid and reliable measure of HR-QoL following ICU admission (11, 12).
The SF-36 uses 36 items to measure eight QoL domains: physical functioning, role limitations due to physical problems, bodily pain, general health perceptions, energy/vitality, social functioning, role limitations due to emotional problems, and mental health (13). It has demonstrated acceptability, reliability, and validity in ICU populations and has been validated for self-administration and telephone interview (13–16). The SF-36 has two population-normalized, latent factor based summary scores described as Physical Component Summary (PCS) and Mental Component Summary (MCS) scores with a population mean score of 50 and each 10 point increment (decrement) equal to a SD from the mean (17).
The AQoL is a 15-item generic health, multi-attribute utility instrument comprising four dimensions: independent living, social relationships, physical senses, and psychologic wellbeing. The AQoL produces a utility index that ranges from –0.04 (worst possible QoL) to 1.00 (full QoL). The AQoL has been validated for use in ICU populations (18).
HR-QoL was assessed at baseline, 6 months and 1 year following randomization. For the baseline assessment, the participant’s next of kin was asked to complete the HR-QoL questionnaire (postrandomization, while the participant remained in ICU or the ward) based on the participant’s HR-QoL before their current acute illness (14).
For participants recruited in Australia and New Zealand, site research staff sent a HR-QoL referral form to the trial-coordinating center after the 90-day survival assessment had been completed. Just prior to 6 months following randomization, the trial co-coordinating center mailed HR-QoL questionnaires to all participants known to be alive following their 90-day assessment. Where participants did not return the questionnaires, follow-up contact via telephone was conducted by a central, blinded assessor with participants given the option to provide answers to the EQ-5D-3L over the phone, and instructed to return the completed questionnaires. Similar methods were used just prior to 12 months following randomization, in participants known to be alive at the prior assessment point.
For participants recruited in Finland, Hong Kong, and Ireland, the appropriate translation/adaptation of the EQ-5D-3L was used. In these countries, only the EQ-5D-3L was used to assess HR-QoL due to resource limitations, and due to language differences, 6-month and 1-year assessments were conducted over the phone by site research staff.
Data are presented as proportions for categorical data, and means and SDs or medians and interquartile ranges for continuous data as appropriate. Data were analyzed using Stata v14.2 (StataCorp, College Station, TX). Where data were missing, the number of available observations is reported, and no assumptions were made about missing data. Comparisons between groups were conducted using chi-square tests for binary outcomes and t tests for continuous data. Paired t tests were used for comparing changes between time points within groups. All tests were two-sided with a significance level of 5%.
As the EQ-5D-3L was completed by participants in all participating sites (in five countries), utility scores were valued using the U.K. time-trade-off tariff (19). SF-36 data were analyzed using standard analytic techniques, with scores calculated for each domain as well as calculation of PCS and MCS scores. The AQoL data were analyzed using standard techniques, with calculation of utility scores (20). Domain and summary scores for each HR-QoL instrument were compared with published normative data using t tests (17, 21–23). The minimal clinically important difference (MCID) for each instrument was used to determine the clinical importance of changes in HR-QoL scores between groups, over time and in comparison to population norms. The MCID in SF-36 scores is 2–3 points for the PCS, 3 points for the MCS, and 2–3 points for each domain, other than role limitations due to emotional problems (4 points) (13). The MCID for the EQ-5D-3L utility score is 0.074 (24), the EQ5D VAS is 7 points (25), and the AQoL utility score is 0.06 (21).
Survival curves were estimated using the Kaplan-Meier method, and the log-rank test was used to assess the difference between survival curves for the EGDT and UC groups. A Cox’s proportional hazards model was used to estimate the effect of participant characteristics upon duration of survival, using the same predefined baseline covariates as in the primary article—country, age, Acute Physiology And Chronic Health Evaluation (APACHE) II score, systolic blood pressure (< 90 mm Hg or ≥ 90 mm Hg), and presence/absence of mechanical ventilation. For participants lost to follow-up at 1 year, data were censored at the date the participant was last known to be alive.
A logistic regression model was used to estimate the effect of participant characteristics on 12-month mortality, whereas a linear regression model was used to estimate the effects of HR-QoL using the EQ-5D-3L utility score. Both models used the same baseline covariates as in the primary article. As the response to interventions may depend on baseline function (26), mortality and HR-QoL results were also stratified by the baseline Charlson comorbidity score (with groups defined by a score of 0, 1, or ≥ 2).
From October 2008 until April 2014, 1,600 participants were randomized to receive EGDT (n = 796) or UC (n = 804). Figure 1 shows the flow of participants through the time points. A total of 1,548 (97.3%) and 1,515 (95.2%) participants were followed for survival at 6 and 12 months respectively; a total of 1,013 (84.1%) and 940 (85.1%) surviving participants completed HR-QoL assessments at 6 and 12 months, respectively.
Baseline characteristics for all ARISE participants have been previously reported (6). Table 1 presents baseline characteristics, including baseline HR-QoL (for more detailed HR-QoL at baseline, see Tables S3 and S4, Supplemental Digital Content 2, http://links.lww.com/CCM/E510). There was no difference in HR-QoL before the sepsis episode between groups, other than for the SF-36 mental health domain in which the EGDT group had higher scores than the UC group, although the mean difference of 2 was smaller than the MCID (Table S4, Supplemental Digital Content 2, http://links.lww.com/CCM/E510). HR-QoL in both groups at baseline was significantly lower than population norms for all SF-36 domains and summary scores (p < 0.0001 for all), the EQ VAS (p < 0.0001), the EQ-5D-3L derived utility score (p < 0.0001), and the AQoL utility score (p < 0.0001), with all differences greater than the MCID.
Baseline HR-QoL measures were completed for 50.5% of participants (803/1,591). Nonresponders (i.e., participants for whom HR-QoL measures were not completed) at baseline were more likely to be female, have a higher illness severity and have comorbidities (Table S1, Supplemental Digital Content 2, http://links.lww.com/CCM/E510). Nonresponders at baseline had a significantly higher 90-day mortality than responders (27.6% vs 9.7%; p < 0.001). Nonresponders (excluding deceased participants) at both 6 and 12 months did not differ in any baseline characteristics, except that nonresponders were significantly younger than responders (Table S2, Supplemental Digital Content 2, http://links.lww.com/CCM/E510).
The 90-day, 6-month, and 1-year mortality are presented in Table 2. There were no significant differences between groups at any time point (6 mo EGDT 21.8% vs UC 22.6%; odds ratio (OR), 0.95; 95% CI, 0.75–1.21; p = 0.70 and 12 mo EGDT 26.4% vs UC 27.9%; OR, 0.93; 95% CI, 0.74–1.16; p = 0.50). Figure 2 shows Kaplan-Meier curves for duration of survival between randomization and 1 year, with no difference between groups (p = 0.48). In the Cox regression model, independent predictors of death were increasing age and illness severity, measured using the APACHE II, while treatment group was not significantly associated with duration of survival (Table S7, Supplemental Digital Content 2, http://links.lww.com/CCM/E510). Results were similar in the logistic regression model with no association between treatment group and 12-month mortality (Table S8, Supplemental Digital Content 2, http://links.lww.com/CCM/E510). The lack of association between treatment group and mortality was consistent across the three groups defined by the Charlson comorbidity score (Table S10, Supplemental Digital Content 2, http://links.lww.com/CCM/E510).
Quality of Life
There were no significant differences between groups in HR-QoL as measured by the EQ-5D-3L or the SF-36 at either the 6- or 12-month follow-up (Tables 2 and 3; and Table S5, Supplemental Digital Content 2, http://links.lww.com/CCM/E510). The EQ-5D-3L was completed by 84.1% of surviving participants at 6 months, and 85.1% at 12 months. There were no significant differences between groups in any domain of the EQ-5D-3L at either 6 or 12 months postrandomization, nor was there a difference in the EQ VAS or the utility score (Tables 2 and 3). There were no significant differences between groups in any SF-36 domain at either 6 or 12 months, nor was there any significant difference between groups in the PCS or MCS scores (Table 2; and Table S5, Supplemental Digital Content 2, http://links.lww.com/CCM/E510). Similarly, there were no significant differences in the AQoL utility score at either 6 or 12 months (Table 2). All differences were smaller than the MCID for each HR-QoL measure. Results for all HR-QoL measures were consistent across the 3 Charlson comorbidity score groups (Table S10 in Supplemental Digital Content 2, http://links.lww.com/CCM/E510). In the linear regression model, the only independent predictor of HR-QoL at 12 months (using the EQ-5D-3L utility score) was age, with no significant association between treatment group and HR-QoL (Table S9, Supplemental Digital Content 2, http://links.lww.com/CCM/E510).
All domains of the EQ-5D-3L, the EQ VAS, and the utility scores were significantly lower in both groups compared with age-matched population norms at all time points (all p < 0.0001), with all differences greater than the MCID. Similarly, all SF-36 domains and the PCS and MCS scores were significantly lower than age-matched population norms for both groups at all time points, with all differences greater than the MCID, other than the bodily pain and mental health domains for both groups (Table 2; and Table S5, Supplemental Digital Content 2, http://links.lww.com/CCM/E510). The AQoL utility scores were significantly lower than age-matched populations norms for both groups at all time points (all p < 0.0001), with all differences greater than the MCID (Table 2).
At 12 months, EQ VAS scores were significantly higher than baseline levels in both the EGDT (p = 0.01) and UC (p < 0.0001) groups, although the difference of 3.5 points in the EGDT group was lower than the MCID of 7 (Table S6, Supplemental Digital Content 2, http://links.lww.com/CCM/E510). There was no difference in EQ-5D-3L utility scores. There was no difference in SF-36 PCS scores at 12 months compared with baseline, and although the MCS score in the EGDT group was significantly lower than baseline scores (p = 0.01), the difference of 2.75 points was smaller than the MCID of 3. Of the SF-36 domains, role physical (p = 0.002 for EGDT and p = 0.02 for UC) and bodily pain (p = 0.006 for EGDT and p = 0.04 for UC) domains were lower at 12 months than baseline in both groups although only the difference in role physical was greater than the MCID, indicating participants had not returned to premorbid levels. For the UC group, all other SF-36 domains at 12 months did not differ from baseline values, and although in the EGDT group, role limitation due to emotional problems (p = 0.009) and mental health (p = 0.002) domains were significantly lower than baseline, these differences were smaller than the MCID (Table S6, Supplemental Digital Content 2, http://links.lww.com/CCM/E510). As with the EQ-5D-5L utility scores, there was no difference in AQoL utility scores between baseline and 12 months, indicating participants had returned to their premorbid level.
In this long-term follow-up of a multicenter randomized trial of EGDT versus UC for patients presenting to the ED with early septic shock, we reported patient mortality to 12 months. We found no difference in mortality for patients treated with EGDT or UC. There was also no difference in HR-QoL between EGDT and UC groups, with both groups having HR-QoL scores significantly lower than reference populations at both baseline and 12 months.
Relationship to Previous Studies
Our findings agree with those from other multicenter RCTs of EGDT that EGDT provides no long-term survival benefit over UC. The Protocolized Care for Early Septic Shock (ProCESS) RCT of EGDT conducted in 31 U.S. sites found no difference between protocol-based EGDT, protocol-based standard therapy, or UC to 1-year postrandomization (27). Similarly, the Protocolised Management in Sepsis (ProMISe) study conducted in 56 U.K. hospitals found no difference in mortality at 12 months. Importantly, however, the mortality rate in both studies was significantly higher than seen in the present study (28).
The ProMISe study reported utility scores derived from the EQ-5D-3L at 90 days and 12 months postrandomization and, as in our study, found no difference between groups at either time point (28). Our HR-QoL scores at 12 months are similar to those seen in other studies of patients with severe sepsis or septic shock. A systematic review of QoL after severe sepsis found that critically ill patients had lower HR-QoL than age- and gender-matched populations (5). Few studies have measured HR-QoL premorbidly in patients with sepsis but, consistent with our results, they have shown reduced HR-QoL compared with population norms (29, 30).
Although participants’ HR-QoL scores remained below population norms at 6 and 12 months postbaseline, they had returned to premorbid function 12 months after the early septic shock episode for most HR-QoL measures. In a small Dutch study of 170 patients with severe sepsis, physical functioning, role physical, and general health SF-36 domains at 6 months were still significantly lower than preadmission values (30). At 12 months in our study, both the EGDT and UC groups remained below premorbid levels for the role physical domain but had returned to premorbid levels for physical functioning and general health. These differences between studies may be due to sample size, with the Dutch study evaluating only 95 patients at 6 months postsepsis, compared with 436 patients with completed SF-36 questionnaires in the present study. It may also be due to type 1 errors, with no adjustment for multiple tests, or differences in patient cohorts, location and treatment patterns, or differences in timing of measurements.
In the present study, both groups reported higher overall health (measured using the EQ VAS scores) at 12 months compared with baseline (although only the UC group difference was greater than the MCID), despite no difference in utility scores. It is possible that this change relates to a response shift, described by Hofhuis et al (30) as a change in self-evaluation resulting from alterations in internal standards or values in patients confronted with life-threatening illness. The comparison between time points should be interpreted cautiously given that baseline responses were completed by proxies and are not directly comparable to patients’ responses at 12 months. However, previous research demonstrated that, although proxies tend to attenuate patient ratings, the difference in responses between patients and proxies is not clinically important (31, 32).
The current study is the largest randomized study of EGDT to date and has a high follow-up rate of more than 95% for survival outcomes and more than 80% for HR-QoL outcomes. Previous studies examining the effect of EGDT on QoL had lower follow-up rates or have not reported HR-QoL (27, 28). This is also the only study of EGDT that examined HR-QoL prior to randomization, enabling comparisons of long-term outcomes to baseline. However, the proportion of missing HR-QoL data at baseline was high, limiting the strength of associations between long-term outcomes and baseline HR-QoL. Although the study protocol did not specify a Bonferroni adjustment for multiple comparisons, a post hoc adjustment for the 82 comparisons undertaken for HR-QoL outcomes between groups, between time points and with population norms yielded a Bonferroni adjusted p value of 0.0006. Using this conservative p value, all comparisons with population norms remained significant. However, all between group and between time point HR-QoL comparisons were no longer significant, with the exception that the EQ VAS score at 12 months, which remained significantly higher than the baseline level in the UC group (p < 0.0001).
Among survivors, responders were significantly older than nonresponders. However, this difference does not impact the comparison between EGDT and UC, as there were no differences in responders’ characteristics between the two groups. It is possible that HR-QoL reported in this study was lowered by the higher age of responders compared with nonresponders, although the impact is likely to be minimal given the high response rate.
Our multicenter, randomized trial found no long-term survival benefit and no HR-QoL benefit of EGDT compared with UC for patients with early septic shock. This patient group has significantly lower HR-QoL compared with reference populations, which persists to 12 months following the septic shock episode.
1. Jawad I, Lukšić I, Rafnsson SB. Assessing available information on the burden of sepsis
: Global estimates of incidence, prevalence and mortality
. J Glob Health 2012; 2:010404
2. Brun-Buisson C, Meshaka P, Pinton P, et al; EPISEPSIS Study Group: EPISEPSIS: A reappraisal of the epidemiology and outcome of severe sepsis
in French intensive care units. Intensive Care Med 2004; 30:580–588
3. Gaieski DF, Edwards JM, Kallan MJ, et al. Benchmarking the incidence and mortality
of severe sepsis
in the United States. Crit Care Med 2013; 41:1167–1174
4. Kaukonen KM, Bailey M, Suzuki S, et al. Mortality
related to severe sepsis
and septic shock
among critically ill patients in Australia and New Zealand, 2000-2012. JAMA 2014; 311:1308–1316
5. Winters BD, Eberlein M, Leung J, et al. Long-term mortality
and quality of life
: A systematic review. Crit Care Med 2010; 38:1276–1283
6. Peake SL, Delaney A, Bellomo R; ARISE Investigators: Goal-directed resuscitation in septic shock
. N Engl J Med 2015; 372:190–191
7. Delaney AP, Peake SL, Bellomo R, et al; ARISE investigators: The Australasian Resuscitation in Sepsis
Evaluation (ARISE) trial statistical analysis plan. Crit Care Resusc 2013; 15:162–171
8. Bone RC, Balk RA, Cerra FB, et al. Definitions for sepsis
and organ failure and guidelines for the use of innovative therapies in sepsis
. The ACCP/SCCM Consensus Conference Committee. American College of Chest Physicians/Society of Critical Care Medicine. Chest 1992; 101:1644–1655
9. Angus DC, Carlet J; 2002 Brussels Roundtable Participants: Surviving intensive care: A report from the 2002 Brussels Roundtable. Intensive Care Med 2003; 29:368–377
10. EuroQol G: EuroQol -a new facility for the measurement of health-related quality of life
. Health Policy 1990; 16:199–208
11. Kaarlola A, Tallgren M, Pettilä V. Long-term survival, quality of life
, and quality-adjusted life-years among critically ill elderly patients. Crit Care Med 2006; 34:2120–2126
12. Badia X, Díaz-Prieto A, Rué M, et al. Measuring health and health state preferences among critically ill patients. Intensive Care Med 1996; 22:1379–1384
13. Ware J, Kosinski M, Bjorner J, et al. User’s Manual for the SF-36V2 Health Survey. 2008Second Edition. Lincoln, United Kingdom, QualityMetric Incorporated.
14. Cuthbertson BH, Elders A, Hall S, et al; Scottish Critical Care Trials Group; Scottish Intensive Care Society Audit Group: Mortality
and quality of life
in the five years after severe sepsis
. Crit Care 2013; 17:R70
15. Heyland DK, Hopman W, Coo H, et al. Long-term health-related quality of life
in survivors of sepsis
. Short Form 36: A valid and reliable measure of health-related quality of life
. Crit Care Med 2000; 28:3599–3605
16. Chrispin PS, Scotton H, Rogers J, et al. Short Form 36 in the intensive care unit: Assessment of acceptability, reliability and validity of the questionnaire. Anaesthesia 1997; 52:15–23
17. Hawthorne G, Osborne RH, Taylor A, et al. The SF36 Version 2: Critical analyses of population weights, scoring algorithms and population norms. Qual Life Res 2007; 16:661–673
18. Skinner EH, Denehy L, Warrillow S, et al. Comparison of the measurement properties of the AQoL and SF-6D in critical illness. Crit Care Resusc 2013; 15:205–212
19. Dolan P, Gudex C, Kind P, et al. A Social Tariff for EuroQol: Results from a UK General Population Survey. 1995York, UK, University of York, Centre for Health Economics.
20. Hawthorne G, Richardson J, Day N, et al; Construction and utility scaling of the Assessment of Quality of Life
(AQoL) instrument: Centre for Health Program Evaluation, Monash University, 2000. Report No. Working paper 101
21. Hawthorne G, Osborne R. Population norms and meaningful differences for the Assessment of Quality of Life
(AQoL) measure. Aust N Z J Public Health 2005; 29:136–142
22. Clemens S, Begum N, Harper C, et al. A comparison of EQ-5D-3L population norms in Queensland, Australia, estimated using utility value sets from Australia, the UK and USA. Qual Life Res 2014; 23:2375–2381
23. Szende A, Janssen B, Cabases J. Self-Reported Population Health: An International Perspective Based on EQ-5D. 2014Dordrecht, The Netherlands, Springer.
24. Walters SJ, Brazier JE. Comparison of the minimally important difference for two health state utility measures: EQ-5D and SF-6D. Qual Life Res 2005; 14:1523–1532
25. Pickard AS, Neary MP, Cella D. Estimation of minimally important differences in EQ-5D utility and VAS scores in cancer. Health Qual Life Outcomes 2007; 5:70
26. Puthucheary ZA, Denehy L. Exercise interventions in critical illness survivors: Understanding inclusion and stratification criteria. Am J Respir Crit Care Med 2015; 191:1464–1467
27. Yealy DM, Kellum JA, Huang DT, et al; ProCESS Investigators: A randomized trial of protocol-based care for early septic shock
. N Engl J Med 2014; 370:1683–1693
28. Mouncey PR, Osborn TM, Power GS, et al; ProMISe Trial Investigators: Trial of early, goal-directed resuscitation for septic shock
. N Engl J Med 2015; 372:1301–1311
29. Karlsson S, Ruokonen E, Varpula T, et al; Finnsepsis Study Group: Long-term outcome and quality-adjusted life years after severe sepsis
. Crit Care Med 2009; 37:1268–1274
30. Hofhuis JG, Spronk PE, van Stel HF, et al. The impact of severe sepsis
on health-related quality of life
: A long-term follow-up study. Anesth Analg 2008; 107:1957–1964
31. Dinglas VD, Gifford JM, Husain N, et al. Quality of life
before intensive care using EQ-5D: Patient versus proxy responses. Crit Care Med 2013; 41:9–14
32. Hofhuis J, Hautvast JLA, Schrijvers AJP, et al. Quality of life
on admission to the intensive care: Can we query the relatives? Intensive Care Med 2003; 29:974–979
early goal-directed therapy; mortality; quality of life; sepsis; shock; septic; survival rate
Supplemental Digital Content
Copyright © by 2019 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.