Sex Differences in Long-Term Quality of Life Among Survivors After Stroke in the INSTRUCT

Hoang T. Phan, PhD; Christopher L. Blizzard, PhD; Mathew J. Reeves, PhD; Amanda G. Thrift, PhD; Dominique A. Cadilhac, PhD; Jonathan Sturm, PhD; Emma Heeley, PhD; Petr Otahal, GDipSc; Peter Rothwell, PhD; Craig S. Anderson, PhD; Priya Parmar, PhD; Rita Krishnamurthi, PhD; Suzanne Barker-Collo, PhD; Valery Feigin, PhD; Seana Gall, PhD


Stroke. 2019;50(9):2299-2306. 

In This Article


Qualified investigators can request access to patient-level data, analytic methods, and study materials after ethics clearance and approval by all authors.

The INSTRUCT (International Stroke Outcomes Study) is an international collaboration including individual participant data of n=16 964 people with first-ever stroke from 13 different studies that adhered to the criteria for ideal population-based stroke incidence studies.[5,7] The studies were conducted in Australasia, Europe, South America, and the Caribbean (1987–2014). This INSTRUCT was approved by the Tasmanian Health and Medical Human Research Ethics Committee (H0014861). All the participating studies had approval from their respective local Ethics Committees.

This analysis of INSTRUCT includes 4 studies: Oxford (United Kingdom), Perth (Australia), Melbourne (Australia), and Auckland (New Zealand). Of these, 2 had measures of HRQoL among survivors at both 1 and 5 years (Oxford, Melbourne) while the others had only 1-year (Perth) or 5-year measures (Auckland).

Outcome Measurements

Participants of the studies were followed-up with face-to-face interviews conducted at 1 and 5 years after stroke.[7] Three instruments were used to assess HRQoL. In the Oxford study, HRQoL was assessed using the European Quality of Life-5 Dimensions (EQ5D) instrument[8] which evaluates 5 dimensions: mobility, self-care, usual activities, depression/anxiety, and pain. In the Melbourne study, the Assessment of Quality of Life (AQoL)[9] was used comprising 5 dimensions: illness, independent living, social relationships, physical senses, and psychological well-being. The Short Form-36 (SF36)[10] was used in the Perth and Auckland studies. This instrument has 8 subdimensions: vitality, physical functioning, bodily pain, general health perceptions, physical role functioning, emotional role functioning, social role functioning, and mental health.

HRQoL utility scores were calculated from the 3 instruments among survivors by summing the component scores and adding value sets from relevant populations to generate the utility score. The utility score ranges from full health, with a value of one, to negative values indicating health states worse than death. The AQoL utility scores (Melbourne) were calculated based on the Australian population-based methods[9] while the EQ5D scores (Oxford) were derived from patient data using the available value sets for the UK population.[11] The Short-Form 6-Dimension (SF6D) utility scores were derived from SF36 items (Perth, Auckland) using the published algorithm for the UK population[12] given the lack of SF6D preference weights for Australia or New Zealand. The SF6D is developed for use in economic evaluation by reducing the SF36 to a 6-dimension classification and creating an overall index (utility) for health.[12]

Study Factors

We included covariates that we hypothesized from our previous studies[6,13] might explain sex differences in HRQoL outcomes. These included sociodemographics (n=4 variables), prestroke health (n=13; eg, prestroke dependence; health behaviors/comorbidities), stroke-related factors (n=2), treatment and management (n=14), and poststroke factors (n=3). The availability and specification of individual variables for each study are provided in Supplement I in the online-only Data Supplement.

Statistical Analysis

Data were analyzed using Stata 12.1 (StataCorp Texas, 2011), with 2-tailed P≤0.05 considered statistically significant.

Because of the skewed distribution of the outcome, quantile regression was used to estimate the median difference for women compared with men in utility scores from HRQoL measures among survivors at 1 and 5 years. We used a 2-stage analysis method.[14] Study-specific models for the sex difference in HRQoL were built in the first stage. Within each study, the role of study factors as potential covariates of the association between sex and HRQoL were assessed. Variables were entered into the multivariable models if they met the following criteria: associated with sex, associated with stroke severity, and the inclusion of the covariate changed the magnitude of the sex coefficient by ≥10% (Supplement II in the online-only Data Supplement).[5,7] Where possible, age, stroke severity, and prestroke dependency, variables that are common predictors of stroke outcome and associated with sex,[6,13] were forced into the final multivariable models regardless of meeting the above criteria. Within each study, statistical interactions were assessed by a test of statistical significance of a sex × covariate product term. In the second stage of the analysis, the effect estimates from unadjusted and multivariable-adjusted models from the individual studies were combined to create pooled estimates using random-effects meta-analysis. Statistical heterogeneity between studies was evaluated using Q and I2 statistics. The potential sources of between-study heterogeneity were not assessed because the number of studies forming our pooled estimates was less than required (≥10).[15]

We found substantial disparities in the definitions of the minimum clinically important difference across instruments.[16,17] The comparability of the study populations with measured HRQoL was undertaken to see whether the sex differences in utility scores might be clinically meaningful. In these analyses, the clinically relevant thresholds were 0.06 for AQoL,[16] 0.08 to 0.12 for EQ5D,[17] and 0.03 to 0.08 for SF6D.[17]

Sensitivity Analyses

HRQoL was highly correlated with poststroke functional outcome or activity limitation (correlation coefficient; r=0.5–0.8; Table I in the online-only Data Supplement) but not with mood disorder (r=0.1–0.3). To avoid over adjustment, the functional outcome, assessed using the Barthel Index (Melbourne) or modified Rankin Scale score (remaining studies; Supplement I in the online-only Data Supplement), was not included in the final models. We report further adjustment for this factor in separate models.

For the studies with >20% missing data on HRQoL, assuming that the data were missing at random, we performed multiple imputation combined with inverse probability weighting (Supplement II in the online-only Data Supplement). We compared these results to those from complete-case analyses.

Subdomain Analyses

We examined sex differences in subdomain scores among the Auckland (SF36), Perth (SF36), and Melbourne (AQoL) studies using quantile regression.

Analyses of the Comparison With General Populations

We examined HRQoL between men and women with stroke and men and women in the general populations using linear regression (Supplement II in the online-only Data Supplement).