Very Low Quality of Life After Acute Stroke
Data From the Efficacy of Nitric Oxide in Stroke Trial
Background and Purpose—Health-related quality of life, a key outcome after stroke, plays a role in the analysis of treatment cost-effectiveness. Some measures of health-related quality of life allow for a quality of life worse than death; the characteristics of such patients have not been well described.
Methods—Data from the Efficacy of Nitric Oxide in Stroke (ENOS) trial were used to explore health-related quality of life after stroke. The EuroQol questionnaire (EQ-5D) was performed at day 90, and a health utility score (HUS) was calculated. HUS was defined as follows: poor-good HUS >0, death HUS=0, and very poor HUS <0. The characteristics and outcomes of patients with HUS <0 were then explored.
Results—Of the 2569 patients, 303 (11.8%) died, and of the 2238 with quality of life data available, of whom 724 (32.3%) were completed by a proxy, 1959 (87.5%) had an HUS >0 and 279 (12.5%) had an HUS <0. Patients with HUS <0 were more likely to be older, women, have severe stroke, have proxy responders, and be institutionalized. Dominant hemisphere strokes were more likely to have proxy responders but not HUS <0. HUS was strongly correlated with dependency (modified Rankin Scale, r=−0.78) and disability (Barthel index, r=0.84) and moderately correlated with mood (Zung depression score, r=−0.59) and baseline severity (r=0.51). All but 1 patient with modified Rankin Scale of 5 had an HUS <0.
Conclusions—Very low health-related quality of life is relatively common after stroke, particularly in patients with mobility problems or who are dependent on help for usual activities, and is related to poor functional outcome measures.
Health-related quality of life (HRQOL) is reduced in stroke survivors and is now a commonly used outcome measure in stroke trials, with particular importance to patients.1 Broadly, HRQOL after stroke is related to functional outcome. Despite this, there are variations that cannot be explained by function alone. For example, HRQOL differs between sexes, with women having worse HRQOL after stroke, even after correction for age and prognostic factors.2 Similarly, even after correction for outcome, HRQOL varies between countries, with stroke patients in different countries rating physical and mental health domains in varying manners.3
In addition to providing patients’ perspective of life after stroke, HRQOL scales are used in health economics to assess cost-effectiveness of treatments. Embedded in these analyses is the principle that in some instances, health states may be considered as worse than death.4 The characteristics of stroke survivors who have very low HRQOL scores, equivalent to health states worse than death, have not been explored in the literature.
Using data from the large ongoing international Efficacy of Nitric Oxide in Stroke (ENOS) trial,5 we sought to examine patients with very poor HRQOL. We aim to describe patients with very low HRQOL scores, including baseline factors and functional outcome, to determine characteristics associated with a health state worse than death.
Patients Included in Analysis
Data were obtained on patients with acute stroke enrolled in the ongoing ENOS trial between the period of July 2001 and March 2012. ENOS is a prospective international multicenter randomized controlled trial testing transdermal glyceryl trinitrate versus control in acute (<48 hours) stroke5; patients taking antihypertensive agents before their stroke are also randomized to continue or stop this temporarily. Treatment is given for 7 days (www.enos.ac.uk/). Key baseline inclusion criteria include ischemic stroke or primary intracerebral hemorrhage, systolic blood pressure >140 mm Hg, motor impairment, and premorbid modified Rankin Scale (mRS) <3. The primary outcome is shift in the mRS at day 90.6 All information is collected prospectively as part of the trial protocol. The trial is being conducted according to the Declaration of Helsinki and the International Conference on Harmonization of Good Clinical Practice.
Health-Related Quality of Life
HRQOL was assessed at day 90 by telephone interview using the EuroQol, which consists of EQ-5D and visual analog scale (VAS).7 The EQ-5D examines the respondent’s view on mobility, self-care, usual activities, pain or discomfort, and anxiety or depression as part of a total health state. Each part is awarded ≤3 points, the score representing no problem, some problem, or extreme problem (1, 2, or 3, respectively). The scores are then combined to form a 5-digit health state, which is then converted into a health utility score (HUS) using the time trade-off algorithm for the United Kingdom.8 An HUS <0 represents very poor HRQOL and is judged to represent a health state worse than death, with death having a value of 0.8 In the final part of the EuroQol (ie, the VAS), the respondents are asked to mark their current health state on scale of 0 (worst imaginable health state) and 100 (best imaginable health state).9 Where patients were unable to complete the outcome questions themselves, carers were permitted to respond on their behalf.
Functional outcome was assessed as dependency (mRS, ENOS primary outcome), disability (Barthel index [BI]), mood (Zung depression scale),10 cognition (modified telephone mini-mental state examination), and disposition (home/institution), each determined at day 90 via a structured telephone interview blinded to treatment assignment. Patients were excluded from those parts of the analysis where an assessment could not be completed in full.
Case Mix/Prognostic Factors
Case mix variables included demographics (age, sex, geographical region), vascular risk factors (hypertension, diabetes mellitus, previous stroke, previous myocardial infarction), premorbid dependency (mRS), stroke severity (Scandinavian Stroke Scale [SSS]),11 stroke syndrome,12 and atrial fibrillation on ECG.
Differences between subgroups were investigated using χ2 test for categorical (unordered) variables or t test for continuous variables. For each EQ-5D domain, unadjusted and adjusted ordinal logistic regression was used to investigate subgroup differences, taking into account the severity (ordered levels) of EQ-5D domains. Unadjusted models included only the subgroup of interest. Adjusted models included the subgroup of interest plus all factors statistically significant in univariate analyses. As the trial is still ongoing, analyses were blinded to treatment allocations (glyceryl trinitrate versus no glyceryl trinitrate, continue versus stop premorbid antihypertensives). All analyses were performed using SAS (version 9.3). Significance was taken at P<0.05, and 95% confidence intervals are given.
Of the 2569 patients with confirmed diagnosis of acute stroke enrolled in the ENOS study, 303 (11.8%) patients died by day 90. EQ-5D was available in 2238 patients, of whom 724 (32.3%) were completed by a carer or other proxy. A total of 279 patients (12.5%) had an HUS <0 (Figure). Patients came from 19 countries across 5 continents. Of the 28 patients with missing QOL data, 6 were lost to follow-up, 2 were missing respondent details, 15 were proxy responders, and 5 were participant responders.
Patients with HUS <0 were more likely to be older, women, have severe stroke, have total anterior circulation stroke syndrome, and be in atrial fibrillation. Previous hypertension, previous myocardial infarction, and prestroke dysfunction (mRS >0) were more likely in the HUS <0 group. Information for those with HUS <0 mostly came from proxy respondents (72.8%). Baseline stroke severity was greatest in patients who died (mean SSS 22, equivalent13 to National Institute of Health Stroke Scale 14) or had very low HRQOL (mean SSS 24, equivalent to National Institute of Health Stroke Scale 13). The distribution of HUS <0 versus HUS ≥0 patients varied by geographical region, with higher percentage of HUS <0 in patients from the British Isles (13.9%) and Europe (15.3%) compared with patients from Asia (4.8%) and North Africa (8.2%; Table 1). Dominant hemisphere strokes were more likely to have proxy responders but not HUS <0 (Table 2). Proxy responders were more likely to have total anterior circulation stroke syndrome.
In contrast to patients with HUS >0, those with HUS <0 had a worse functional outcome (mRS and BI), lower mood, and were more likely to be in an institution or still in hospital (Table 3). HUS was correlated with all the functional outcome measures: mRS (r=−0.78; P<0.001), BI (r=0.84; P<0.001), EQ-VAS (r=0.58; P<0.001), Zung depression scale (r=−0.59; P<0.001), and mini-mental state examination (r=0.32; P<0.001), as well as baseline SSS (r=0.51; P<0.001; Table I in the online-only Data Supplement). All but 1 patient with mRS=5 had an HUS <0. Proxy responders had significantly worse outcomes than participant responders (Table 4).
Domains of EQ-5D
The majority of patients with HUS <0 reported extreme problems in usual activities (95%), mobility (81%), and self-care (83%). In contrast, extreme problems with pain/discomfort (29%) and anxiety/depression (24%) were less common (Table II in the online-only Data Supplement).
Proxy respondents reported less extreme pain/discomfort than self-respondents (22% versus 47%; P=0.0001) but more extreme problems with self-care (90% versus 66%; P<0.0001) and mobility (89% versus 61%; P<0.0001; Figure I in the online-only Data Supplement).
Although ENOS only recruits people who are previously independent (mRS <3), 10% of patients died within 3 months, and of the survivors, 279 (13%) survived with very poor HRQOL (ie, HUS <0).
As expected, baseline predictors (increasing age and severity, female sex, atrial fibrillation, mild premorbid dysfunction) were all related to QOL. It is notable that dominant hemisphere stroke did not independently predict lowest quality of life scores, which is perhaps surprising, especially as a large proportion of these patients had proxy respondents who were likely to over-report rather than under-report problems. This may reflect that the generic EuroQol does not measure the limitations and impairments attributable to problems with a specific hemisphere. An alternative stroke-specific measure, including senses or speech or cognitive function in the QoL instrument, may produce different results.14
As reported previously, functional outcome closely correlated to quality of life scores, with most dependent patients having the lowest scores.15 Problems were most commonly reported with usual activities, self-care, and mobility, with BI the most closely correlated outcome, perhaps as these measures are recording similar activities, explaining why BI has been shown to predict QOL scores.16
The mRS has also been used to reflect HRQOL after stroke, particularly when the possible effect of certain intervention, such as hemicraniectomy, is that patients survive at the expense of severe neurological deficit.17,18 This has been explored in several recent trials, where a bad outcome is classified as an mRS >3. Although our data confirm that patients with an mRS of 4 do have low quality of life scores (mean HUS, 0.3; VAS, 56.5), patients with an mRS of 5 had a mean utility score <0 (mean HUS, −0.1; VAS, 32.3), representing an outcome that the general public may perceive as worse than death.
It is recognized that stroke survivors rate HRQOL higher than the general population, possibly because of the development of coping strategies.19 This is perhaps reflected in the EQ-VAS, which, despite being lower than population values, was higher than that expected compared with the HUS. The minimal clinically significant difference in EQ-VAS has previously been reported as 10; our data reflect that there is a significant difference between HRQOL in those with an mRS of 4 (mean VAS, 56.5) compared with those with an mRS of 5 (mean VAS, 32.3),20 suggesting perhaps that mRS of 4 and 5 should not be considered together as a single outcome.
We used proxy scores when the patients were unable to complete the questionnaire for themselves because such proxy responses had more severe stroke, more frequently affecting the dominant hemisphere. Although proxies are recognized to over-report problems with activities, the difference between proxy and self-respondents is not significant.21 We repeated the analyses excluding proxy responders, and the results remained unchanged, notably dominant hemisphere stroke was not associated with very low QOL and functional outcome was strongly correlated to QOL, even when participant responders only were included. Proxy respondents were less likely to report problems with pain or discomfort and anxiety or depression than nonproxy respondents. The use of proxy scores in patients with stroke has been validated in the literature,22 but correlation with participant scores is only modest and is affected by depression and proxy perception of burden.23
Depression is common after stroke, and the role of depression in reporting QOL scores is well recognized.14 Although there was a mild relationship between QOL and mood, usual activities, mobility, and self-care may play a more important role in determining quality of life scores than the mental health domains. HRQOL was only weakly correlated with cognition, as has been shown previously.24
Several limitations are present in this study. First, QOL scores were reported across many different geographical regions; variations in QOL across regions and cultures are widely recognized after stroke and can have a small but significant effect on reported scores. Second, although patients came from 19 different countries, HUS was calculated using the UK algorithm, reflecting that the majority (63.8%) of patients were recruited from the United Kingdom. Our use of HUS <0 is subjective and reflects the UK sample’s opinion that, faced with the choice of being in certain states or being dead, they would prefer death. However, the number of health states that fall <0 point depends on the population; in the American model, there are a smaller number of domains deemed worse than death.25 Third, the data came from a randomized controlled trial that selected patients with acute stroke on the basis of high blood pressure, motor impairment, and previous independence. Such criteria may have altered both the proportion of patients with very low scores and their demographic and clinical characteristics. Finally, the value of attempting to quantify something that is adjudged by some to be nonexistent (quality of life worse than death) is debatable and worthy of wider discussion. Nevertheless, this large data set comes from a high fidelity trial with prospective data collection, and including negative values (such as life worth than death) in cost-effectiveness analysis removes the possibility of a floor effect, which would underestimate a possible treatment effect.26
In summary, we have demonstrated that very low HRQOL scores are relatively common after stroke, with a proportion of patients recording scores representing an outcome that the population may perceive as worse than death. Patients with limitations in mobility and usual activities were the most likely to have the lowest scores; however, dominant hemisphere stroke was not a predictor of the lowest scores. Patients with an mRS of 5 had a mean HUS <0. Interventions that reduce mortality at the expense of survival with increased dependency may be conversely worsening quality of life, an issue that warrants debate among clinicians, stroke survivors, and their carers.
We thank Efficacy of Nitric Oxide in Stroke (ENOS) investigators for enrolling patients into the ongoing ENOS trial, ENOS Trial Steering Committee for commenting on the manuscript, and Cyrille Correia who assisted with the statistical analysis. P.M.W. Bath is Stroke Association Professor of Stroke Medicine and Chief Investigator of the ENOS trial.
ENOS Trial Steering Committee—Independent Chairman: Graham Venables (Sheffield); Independent experts: Pierre Amarenco (Paris, France), Keith Muir (Glasgow); Members: Philip Bath (Nottingham), Kennedy Lees (Glasgow), Stuart Pocock (London), Joanna Wardlaw (Edinburgh), David Whynes (Nottingham), and Eivind Berge (Oslo, Norway).
Sources of Funding
Efficacy of Nitric Oxide in Stroke is funded by British United Provident Association Foundation, Hypertension Trust, and the Medical Research Council.
Guest Editor for this article was Natan Bornstein, MD.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.113.002201/-/DC1.
- Received May 24, 2013.
- Accepted September 17, 2013.
- © 2013 American Heart Association, Inc.
- Gray LJ,
- Sprigg N,
- Bath PM,
- Boysen G,
- De Deyn PP,
- Leys D,
- et al
- Gray LJ,
- Sprigg N,
- Bath PM,
- Sørensen P,
- Lindenstrøm E,
- Boysen G,
- et al
- Bath PM,
- Lees KR,
- Schellinger PD,
- Altman H,
- Bland M,
- Hogg C,
- et al
- Stiggelbout AM,
- de Haes JC
- Whynes DK,
- Sprigg N,
- Selby J,
- Berge E,
- Bath PM
- Williams LS,
- Bakas T,
- Brizendine E,
- Plue L,
- Tu W,
- Hendrie H,
- et al