The Short-Form Six-Dimension utility index predicted mortality in the European Prospective Investigation into Cancer-Norfolk prospective population-based study

Journal of Clinical Epidemiology 63 (2010) 192e198

The Short-Form Six-Dimension utility index predicted mortality in the European Prospective Investigation into Cancer-Norfolk prospective population-based study Phyo K. Myinta,b,*, Richard D. Smithc, Robert N. Lubend, Paul G. Surteesd, Nicholas W.J. Wainwrightd, Nicholas J. Warehame, Sheila A. Binghamf, Kay-Tee Khawb a

Ageing and Stroke Medicine Section, Health & Social Sciences Research Institute, School of Medicine, Health Policy and Practice, University of East Anglia, Chancellors Drive, Norwich, Norfolk NR4 7TJ, UK b Clinical Gerontology Unit, School of Clinical Medicine, University of Cambridge, Cambridge, UK c Health Policy Unit, Department of Public Health and Policy, London School of Hygiene and Tropical Medicine, London, UK d Strangeways Research Laboratory, Department of Public Health and Primary Care, University of Cambridge, Cambridge, UK e MRC Epidemiology Unit, Cambridge, UK f MRC Centre for Nutritional Epidemiology in Cancer Prevention and Survival, Cambridge, UK Accepted 8 May 2009

Abstract Objective: To examine the relationship between the Short-Form Six-Dimension (SF-6D) and mortality. Study Design and Setting: Participants were 17,736 men and women aged 40e79 years at baseline who lived in Norfolk, UK, and had no known cardiovascular disease or cancer, and completed the anglicized Short-Form 36 (SF-36)-item during 1996e2000 in the European Prospective Investigation into Cancer-Norfolk prospective population study. The SF-36 data were converted to SF-6D. The relationship between SF-6D and all-cause and cause-specific mortality were examined. Results: One thousand and seventy deaths occurred during a total of 115,255 person years of follow-up (mean 6.5 years). Lower SF-6D was associated with increased risk of all-cause mortality in men and women. A decrease of 1 standard deviation (0.12 point) in SF-6D was associated with a 35% increase in all-cause mortality (hazards ratio 5 1.35; 95% CI: 1.26, 1.45) after controlling for age, gender, body mass index, systolic blood pressure, cholesterol, diabetes, smoking, and social class. Similar results were observed for cardiovascular, cancer, and other causes of deaths. Conclusion: Poor health utility measured by the SF-6D predicted increased risk of all-cause and cause-specific mortality in men and women. The present study provides the first evidence of the sensitivity of the SF-6D in predicting mortality in an apparently healthy population. Ó 2010 Elsevier Inc. All rights reserved. Keywords: Short-Form 6-Dimension; UK Short-Form 36; Mortality; Health-related quality of life (HRQoL); Preference-based HRQoL; Utility index

1. Introduction Population health is traditionally measured using objective health outcome measures, such as mortality and morbidity statistics [1]. However, more subjective functional health has been increasingly recognized as an important, meaningful, and valid way of assessing health [2]. One of the most widely used measures of functional health is the Short-Form 36 (SF-36) [2].

* Corresponding author. Ageing and Stroke Medicine Section, Health & Social Sciences Research Institute, School of Medicine, Health Policy and Practice, University of East Anglia, Chancellors Drive, Norwich, Norfolk NR4 7TJ, UK. Tel.: þ44 (0) 1603-591942; fax: þ44 (0) 1603-593752. E-mail address: [email protected] (P.K. Myint). 0895-4356/10/$ e see front matter Ó 2010 Elsevier Inc. All rights reserved. doi: 10.1016/j.jclinepi.2009.05.002

However, measures such as the SF-36 are limited in their ability to be used within economic evaluations. The health status profile that is produced does not allow an unambiguous assessment of the improvement or deterioration of functional health unless all elements of the profile move in the same direction, and it does not provide the single index score required to derive a quality-adjusted life year (QALY) as required for economic evaluation [3,4]. For economic evaluation, a measure is required that assigns preferencedor utilitydweights to the different dimensions within the instrument that reflect their relative value in contributing to functional health and can be summed to generate a single overall index of health status. Although such utility measures have been developed in their own right, with the EuroQol (EQ-5D) being one of

P.K. Myint et al. / Journal of Clinical Epidemiology 63 (2010) 192e198

What is new?  There is an inverse relationship between health utility index, Short-Form Six-Dimension (SF-6D), derived from Short-Form 36, and mortality in a general population.  The relationship between SF-6D and mortality is independent of known risk factors for mortality including age, smoking, social class, and other strong biological determinants such as blood pressure, body mass index, and cholesterol level.  It appears that SF-6D may serve as an additional marker for identifying the most vulnerable population at risk, and their use in appropriate settings will provide useful information with regard to targeted preventive interventions.

the most well known of these, there has been considerable work to map health status profiles, such as SF-36, to a preference-based utility index [5]. Brazier et al. [6] derived such a single index measure from the SF-36; the ShortForm Six-Dimension (SF-6D). The rationale, mathematical model and derivation of the evaluation index have been described comprehensively by Brazier et al. [6,7]. The reliability and validity of the SF-6D in a population such as patients with spinal cord injury [8] and systemic sclerosis [9] have been previously reported. However, Moock and Kohlmann [10] compared different preference-based quality-of-life measures including the SF-6D in a sample of patients with mild-to-moderate disease and found that differences between measures may have considerable effects in health economic evaluation studies, and cautioned that it is essential to select carefully instruments for a given study. Moreover, in a mathematical model constructed by Marra et al. [11], very different incremental costeutility ratios were generated depending on the method for determining utility values used in the calculation of QALYs. Given that the move from the SF-36 profile to a single index figure, SF-6D, requires substantial transformation of the original SF-36 data, one of the key issues with such instruments is the impact upon the instruments sensitivity in predicting mortality. We have previously reported an inverse relationship between the physical and mental component summary scores (PCS and MCS) of the SF-36 and all-cause and cause-specific mortality in the European Prospective Investigation into Cancer (EPIC)-Norfolk prospective population-based study [12,13]. The purpose of the present investigation is therefore to gain a deeper insight into the usefulness of preferencebased utility index SF-6D in predicting mortality in a free-living general population who are free of prevalent major illnesses in the UK, EPIC-Norfolk prospective

193

population study. To our knowledge, this is also the largest mapping of the SF-36 data into SF-6D to examine the relationship with total and cause-specific mortality in an apparently healthy free-living general population.

2. Methods 2.1. Participants The study population consisted of men and women living in the general community aged between 40 and 79 years at the baseline participating in the EPIC-Norfolk. They were recruited between 1993 and 1997 and approximately 30,000 (40% response rate) consented to participate. The Norwich Local Research Ethics Committee approved the study. The primary aim was to examine the relationships between dietary intake and risk of incident cancers in a general population. The secondary aim was to study the relationships between diet, psychosocial, and other lifestyle and disease risk factors and a variety of health outcomes including mortality and chronic conditions. Detailed descriptions of the recruitment and study methodology have been previously reported [14]. Briefly, all eligible individuals in the age range in each participating general practice database were invited by mail. Those who consented to participate were asked to provide baseline survey data and were invited to attend for a health examination. 2.2. Measurements At the baseline assessment in 1993e1997, measures were taken by trained staff according to standardized protocols [15]. Body mass index (BMI) was calculated as weight in kilograms divided by the square of height in meters: weight (kg)/[height (m)]2. Blood pressure was measured using an Accutorr Sphygmomanometer after each participant had been seated resting for 5 minutes. Mean of two measurements of blood pressure was used in analysis. Blood samples were also taken at baseline clinic visit. Serum levels of total cholesterol were measured on fresh samples with the RA 1000 (Bayer Diagnostics, Basingstoke, UK). From responses to the questions ‘‘Have you ever smoked as much as one cigarette a day for as long as a year?’’ and ‘‘Do you smoke cigarettes now?,’’ smoking status was classified as current smoker, former smoker, or those who had never smoked. At baseline, social class was classified according to the Registrar General’s occupation-based classification scheme [16]. Social class I consists of professionals, class II includes managerial and technical occupations, class III is subdivided into nonmanual and manual skilled workers, class IV consists of partly skilled workers, and class V comprises unskilled manual workers [17]. In this study, we used social class obtained at the baseline survey in 1993e1997. With the baseline health questionnaire, the participants were asked, ‘‘Has the doctor ever told you that you have

194

P.K. Myint et al. / Journal of Clinical Epidemiology 63 (2010) 192e198

any of the following?,’’ followed by a list of specific conditions including heart attack, stroke, cancer, and diabetes to obtain self-reported prevalent illnesses. 2.3. Predictor variable (SF-6D) During 1996e2000, the surviving participants, then aged 41e80 years, were asked to complete by mail a detailed Health and Life Experiences Questionnaire (HLEQ), which included the anglicized version of SF-36 (UK SF-36) [18,19]. From the individual items of SF-36, mapping to the SF-6D was performed on all the participants who responded to the HLEQ using the software program (SF-6D UK revised. SPS, Sheffield, UK) provided by the Sheffield Health Economics and Decision Science, University of Sheffield. A total of 19,498 participants out of total of 20,921 (93.2%) who responded to SF-36 had adequate data for this conversion. 2.4. Outcome measures All individuals were flagged for death certification at the UK Office of National Statistics (ONS), with vital status ascertained for the whole cohort. ONS reports deaths in the cohort via a regular record linkage system. The mortality outcome data were collected on yearly basis by regular record linkage with ONS. Causes of death were classified as death due to all causes, underlying cardiovascular disease, cancer, and other (non-cardiovascular, non-cancer deaths) causes. Cardiovascular death was defined as Ninth Revision of the International Classification of Diseases (ICD 9) codes 401e448 and Tenth Revision of the International Classification of Diseases (ICD 10) codes I10eI79, and cancer deaths as ICD 9 codes 140e208 and ICD 10 codes C00eC97. For this study purpose, follow-up for each participant began at the date of completion of the HLEQ (SF-36 completion date). We present results for mortality up to the end of July 2005, approximately 6.5 years of average follow-up from the time of completion of the SF-36 questionnaire. 2.5. Statistical analysis Statistical analyses were performed using SPSS for version 14.0 (SPSS Inc, Chicago, IL). After excluding participants with prevalent cancer, myocardial infarction, and stroke, the participants of the EPIC-Norfolk cohort who had available data on SF-36, which could be imputed for the SF-6D were included in this study. Participants with missing values for covariates used in different models were excluded in individual regression analyses. The distribution of sample characteristics of risk factors was examined by quintiles of SF-6D for men and women separately. Crude mortality rates were calculated for all causes, cardiovascular, cancer, and other causes of deaths in each quintile category. Cox proportional hazards models were constructed to determine the independent contribution of the SF-6D utility index and covariates including age, smoking (current smokers vs. ex/non-smokers), diabetes (yes or no), BMI (continuous),

systolic blood pressure (continuous), cholesterol (continuous), and social class (manual vs. nonmanual) for mortality of all cause, cardiovascular, cancer, and other causes for the whole sample (sex included in adjustment) and in men and women separately. The magnitude of clinical impact was assessed for every 0.12-point decrease (equivalent of 1 standard deviation [SD] decrease) in SF-6D scores. Different levels of adjustment were made to examine the relative contribution of age and other covariates in the model. Analyses were repeated after excluding those who died within 2 years of follow-up after HLEQ questionnaire completion to exclude people who might be most likely to have preclinical serious disease at the time of survey.

3. Results After exclusion of 1,762 participants with prevalent heart attack, stroke, and cancer, a total of 17,736 participants (7,763 men and 9,973 women) were eligible to be included in the study. There were 1,070 deaths over 115,255 total person years of follow-up (mean 6.5 years). Table 1 shows the sex-specific distribution of sample characteristics by quintile categories of SF-6D. Quintile 1 represents the lowest and the quintile 5 represents the highest quintile values of SF-6D. Men and women who were in the higher quintile categories of SF-6D were younger, had lower BMI and systolic blood pressure, and had lower prevalences of current smokers, diabetes, and those in manual social classes. Table 2 shows the crude mortality rate for all causes, cardiovascular, cancer, and other causes (non-cancer, noncardiovascular deaths). There appeared to be a linear and significant inverse relationship between SF-6D and mortality for all causes, cardiovascular, cancer, and other causes of deaths for both men and women. Table 3 shows the hazard ratios (HRs) and 95% confidence intervals (CIs) for increased risk of mortality by 1 SD decrease in SF-6D (0.12) by various levels of adjustments. In the fully adjusted model (model B), a decrease in SF-6D scores by 1 SD was associated with a significant increase in risk of death from all causes, cardiovascular, cancer, and other causes of death in this sample. There was a less consistent relationship between men and women for cancer mortality. Repeating the analyses after excluding early deaths showed no significant relationship for cancer mortality in both men and women. The explanatory variables in the Cox proportional hazards model demonstrated proportional hazards. Figure 1 shows the HRs and corresponding 95% CIs for all-cause mortality by quintiles of SF-6D after adjustments for age, sex, BMI, systolic blood pressure, cholesterol, diabetes mellitus, smoking, and occupational social class. Men and women in the lowest SF-6D quintile values had HR 2.08 (95% CI: 1.63, 2.65) compared with people in the top SF-6D quintile.

P.K. Myint et al. / Journal of Clinical Epidemiology 63 (2010) 192e198

195

Table 1 Distribution of variables by quintiles of SF-6D in 7,763 men and 9,973 women of EPIC-Norfolk SF-6D quintiles

Variables Men Age (baseline)* Age (at the time of SF-36)* BMI (kg/m2)* Systolic blood pressure (mmHg)** Cholesterol (mmol/L)z History of diabetes* Cigarette smoking habit* Never smoked Former smoker Current smoker Occupational social class* Nonmanual Manual

1

2

3

4

5

0.30e0.65

0.66e0.75

0.76e0.81

0.81e0.85

0.86e1.00

n 5 1,295 59.8 (9.4) 62.5 (9.4) 26.9 (3.7) 138 (18) 6.0 (1.1) 52 (4.2)

n 5 1,509 58.9 (9.3) 61.5 (9.3) 26.5 (3.3) 138 (18) 6.1 (1.1) 41 (2.7)

n 5 1,554 58.9 (9.2) 61.5 (9.2) 26.3 (3.1) 137 (17) 6.0 (1.1) 43 (2.8)

n 5 1,676 58.7 (9.1) 61.4 (9.1) 26.3 (3.0) 138 (18) 6.1 (1.1) 37 (2.2)

n 5 1,729 57.1 (8.9) 59.6 (8.9) 26.1 (3.0) 136 (17) 6.0 (1.1) 24 (1.4)

396 (30.8) 695 (54.0) 196 (15.2)

521 (34.1) 811 (54.0) 180 (12.0)

573 (37.0) 846 (54.7) 128 (8.3)

580 (34.8) 923 (55.3) 166 (9.9)

694 (40.4) 835 (48.6) 189 (11.0)

679 (53.7) 586 (46.3)

853 (57.5) 631 (42.5)

960 (62.8) 568 (37.2)

1,007 (60.7) 651 (39.3)

1,069 (62.7) 636 (37.3) n 5 1,482 56.2 (8.9) 58.8 (9.0) 25.1 (3.6) 132 (19) 6.2 (1.2) 12 (0.8)

Women Age (baseline)* Age (at the time of SF-36)* BMI (kg/m2)* Systolic blood pressure (mmHg)** Cholesterol (mmol/L)* History of diabetes** Cigarette smoking habit* Never smoked Former smoker Current smoker

n 5 2,257 58.7 (9.7) 61.3 (9.7) 26.9 (4.9) 134 (19) 6.3 (1.2) 50 (2.2)

n 5 2,238 58.1 (9.2) 60.7 (9.3) 26.2 (4.3) 133 (18) 6.3 (1.2) 33 (1.5)

n 5 2,156 58.0 (9.0) 60.6 (9.0) 26.0 (4.1) 134 (18) 6.3 (1.2) 27 (1.3)

n 5 1,840 57.4 (8.9) 60.0 (8.9) 25.7 (4.0) 133 (18) 6.2 (1.2) 17 (0.9)

1,169 (52.3) 735 (32.9) 333 (14.9)

1,287 (57.9) 704 (31.7) 231 (10.4)

1,278 (59.7) 670 (31.3) 191 (8.9)

1,110 (60.9) 550 (30.2) 163 (8.9)

952 (64.7) 392 (26.6) 127 (8.6)

Occupational social class* Nonmanual Manual

1,249 (57.2) 934 (42.8)

1,350 (61.8) 836 (38.2)

1,359 (64.1) 761 (35.9)

1,178 (64.9) 636 (35.1)

946 (65.1) 507 (34.9)

Values are mean (SD) for continuous variables and number (%) for categorical variables. *P ! 0.0001; **P ! 0.01; yP ! 0.05; zP O 0.05.

4. Discussion Men and women with lower SF-6D utility index values had an increased risk of death from all causes, cardiovascular, cancer, and other causes of death after controlling for known risk factors in a general population who were free of prevalent myocardial infarction, cancer, and stroke. Although the associations appeared to be consistent for total and cause-specific mortality, the relationship of SF-6D and cancer mortality appeared to be weaker compared with cardiovascular and other causes of deaths excluding cancer. The inverse relationship between SF-6D score and causespecific mortality was more consistent than either SF-36 physical or mental functional health alone [12,13,20]. It may be that SF-6D being a preference-based measure, takes into account some of the psychological or behavioral aspects of respondents. It is reasonable and plausible that the utility index measure (SF-6D), which can be used in QALY calculations, is strongly related to subsequent mortality outcome. It is intriguing that this relationship is independent of known risk factors for mortality including age, smoking, social class,

and other strong biological determinants such as blood pressure, BMI, and cholesterol level. Moreover, the inverse significant associations were observed across the distribution of sample population spanning different age ranges. In our previous reports [12,13], the particular emphasis was on the relationship between individual major component of the SF-36, physical and mental functional health, and mortality. We addressed the plausible explanations for self-reported health as a determinant of the mortality outcome. As previously discussed, a decline in health may be associated with subtle changes, which cannot be captured by questions about prevalent disease or symptomatic episode, for example angina or transient ischemic attack. Idler et al. have previously described their theories on how self-rated health may predict mortality [21]. The relationship between SF-6D and cancer mortality appeared to be relatively weaker compared with cardiovascular and other causes of deaths excluding cancer. In fully adjusted models (model B, Table 3), like-to-like comparison between SF-36 PCS, SF-36 MCS, and SF-6D showed expected differences [20]. For example, the sex-combined all-cause mortality relative risks for 1 SD decrease in

196

P.K. Myint et al. / Journal of Clinical Epidemiology 63 (2010) 192e198

Table 2 Crude mortality rate (number of deaths) by quintiles of SF-6D in men and women of EPIC-Norfolk SF-6D quintiles 1

2

3

4

5

Cause of death

0.30e0.65

0.66e0.75

0.76e0.81

0.81e0.85

0.86e1.00

All cause Men* Women*

2.6 (200) 1.6 (160)

1.9 (145) 1.0 (102)

1.4 (108) 0.8 (75)

1.3 (104) 0.5 (51)

1.0 (81) 0.4 (44)

Cardiovascular Men* Women*

0.9 (73) 0.6 (57)

0.6 (50) 0.3 (31)

0.5 (40) 0.3 (29)

0.5 (36) 0.1 (13)

0.4 (31) 0.1 (13)

Cancer Men** Womeny

0.8 (61) 0.6 (58)

0.8 (62) 0.6 (55)

0.6 (50) 0.3 (34)

0.6 (49) 0.3 (27)

0.5 (41) 0.3 (26)

Other causes Men* Women*

0.9 (66) 0.5 (45)

0.4 (33) 0.2 (16)

0.2 (18) 0.1 (12)

0.2 (19) 0.1 (11)

0.1 (9) 0.1 (5)

*P ! 0.0001; **P ! 0.01; yP ! 0.05; zP O 0.05.

PCS and MCS were 1.41 and 1.14, respectively; the corresponding HR for 1 SD decrease in SF-6D was 1.35. A similar pattern was observed with regards to cardiovascular and cancer deaths. For other causes of death, the associations appeared to be stronger with SF-6D compared with component summary scores of SF-36. The underlying reason is unclear but it could be attributed by heterogeneous nature of other causes of deaths in a relatively smaller number of outcomes compared with deaths due to cancer and cardiovascular causes. There have been small-scale studies in some selected patient populations examining the relationship between SF6D and mortality [22]. To our knowledge, the present study is the first study to map the SF-36 health profile to the SF-

6D utility index in a free-living population-based cohort to examine mortality with long-term follow-up. Because EPIC-Norfolk is a prospective cohort study with important health outcomes such as cardiovascular and cancer deaths ascertained, we were able to examine the prospective relationships between SF-6D and these objective health outcomes. There was also comparable risk prediction with that of PCS and MCS of the SF-36 health profile. Therefore, our study provides a valuable insight into the convergent validity of these self-rated health measures, as profiles or as a utility index. Furthermore, we have addressed the one of the key issues with such instruments, that is, the impact upon the instruments’ sensitivity in predicting mortality. Our findings provide convincing evidence of sensitivity of the SF-6D in its

Table 3 Number of deaths and adjusted HRs (corresponding 95% CI) for all-cause, cardiovascular, cancer, and other causes of mortality by 0.12-point (1 SD) decrease in SF-6D score in men and women aged 41e80 years in 1996e2005 All

Men

Women

Mortality

N

HR

N

HR

N

HR

All cause A B C

1,070 794 626

1.44 (1.36e1.53) 1.35 (1.26e1.45) 1.27 (1.18e1.38)

638 489 380

1.50 (1.40e1.62) 1.45 (1.33e1.58) 1.33 (1.21e1.47)

432 305 246

1.36 (1.24e1.49) 1.21 (1.08e1.36) 1.18 (1.04e1.34)

Cardiovascular A B C

373 268 200

1.44 (1.31e1.59) 1.37 (1.22e1.55) 1.37 (1.19e1.57)

230 169 121

1.48 (1.31e1.67) 1.44 (1.25e1.67) 1.39 (1.16e1.65)

143 99 79

1.38 (1.18e1.62) 1.25 (1.03e1.52) 1.31 (1.05e1.63)

Cancer A B C

463 363 287

1.22 (1.11e1.33) 1.15 (1.04e1.28) 1.05 (0.93e1.18)

263 209 165

1.25 (1.11e1.41) 1.22 (1.07e1.40) 1.11 (0.94e1.29)

200 154 122

1.18 (1.03e1.35) 1.07 (0.90e1.26) 0.98 (0.81e1.18)

Others A B C

234 163 139

1.96 (1.74e2.21) 1.85 (1.60e2.14) 1.65 (1.40e1.94)

145 111 94

2.09 (1.80e2.42) 1.97 (1.65e2.34) 1.73 (1.43e2.10)

89 52 45

1.76 (1.45e2.15) 1.58 (1.20e2.08) 1.43 (1.07e1.92)

A 5 Adjusted for age (sex for combined analyses), B 5 adjusted for age (sex), BMI, systolic blood pressure, blood cholesterol, diabetes, cigarette smoking, and social class, C 5 same as model B after excluding deaths occurring within 2 years of follow-up using the Cox Proportional Hazards model.

P.K. Myint et al. / Journal of Clinical Epidemiology 63 (2010) 192e198

Fig. 1. HRs and corresponding 95% CIs for all-cause mortality by quintiles of SF-6D in men and women of EPIC-Norfolk adjusted for age, sex, BMI, systolic blood pressure, cholesterol, diabetes mellitus, smoking, and occupational social class (1993/97e2005). Q1 represents the bottom and Q5 represents the top quintile categories of SF-6D.

relationship to all-cause and cause-specific mortality. This is entirely plausible considering the body of existing evidence that the health state of an individual may be modifiable, and health-related behaviors have a substantial impact on mortality [23] and functional health [24e28]. A recent study by Wee et al. [29] demonstrated that in a South-East Asian population, although clinically important differences in utility measurements were present for different preference-based instruments, the impact of these differences on costeutility analyses appeared relatively minor between the EQ-5D, Health Utilities Index Mark 2, and Mark 3 (HUI3), and SF-6D. Chronic medical conditions, marital status, and family functioning influenced the magnitude of these differences. In this study, we excluded people with prevalent illness, and also adjusted for physiological, biological, and social factors. Therefore, our study findings may be applicable to other utility measures. However, in a report by Hatoum et al. [30] comparing the HUI3 and the SF-6D in a multinational phase III clinical trial in patients undergoing percutaneous coronary intervention before hospital discharge and 6 months thereafter, the authors stated that these measures generated different estimates of health state values for their patient population. The authors speculated that these differences might in part be the consequence of the health status descriptive system for the Health Utilities Index that may have been more in line with the hospitalized state than that for the SF-6D. In another study by Brazier et al. [31] where EQ-5D and SF-6D comparison was made across seven patient/population groups, the authors found that there was evidence for floor effects in the SF-6D and ceiling effects in the EQ-5D. They stated that these discrepancies arose from differences in their health state classifications and the methods used to value them. Nevertheless, recently published U.S. population norms from National Health Measurement survey for six health-related quality-of-life utility indices showed similar population trends in U.S. older adults [32]. Therefore, another potential implication of our study findings is their applicability to other health utility

197

measures in predicting mortality. At the moment, it appears that both SF-36 and SF-6D may serve as an additional marker for identifying the most vulnerable population at risk, and their use in appropriate settings will provide useful information with regard to targeted preventive interventions, such as managing modifiable classical risk factors for example blood pressure, and cholesterol level more aggressively. However, we controlled for risk factors including cholesterol and blood pressure and still found a strong relationship between the SF-6D and mortality. It is plausible that the relationships observed are mediated by other factors, which are not controlled in the current report such as anxiety and depression. The nature of these associations requires further exploration to better understand how best to prolong health life expectancy to promote healthy aging.

Acknowledgments We would like to thank participants and general practitioners who took part in the study. We also thank the staff of EPIC-Norfolk and our funders. Funding: EPIC-Norfolk is supported by research program grant funding from Cancer Research UK and the Medical Research Council. The EPIC-Norfolk HLEQ research program is supported by a program grant from the Medical Research Council UK (G0300128). Conflict of interest: None. Ethics approval: Norwich Local Research Ethics Committee approved the study. The corresponding address for the LREC is Clinical Governance Department, Aldwych House, 57 Bethel Street, Norwich. Contributors: K.T.K., S.A.B., and N.J.W. are principal investigators in EPIC-Norfolk population study. P.G.S. is the principal investigator of EPIC-Norfolk HLEQ program. R.N.L. is responsible for data management, computing, and data linkages and P.K.M. conducted analyses. P.K.M. and R.D.S. prepared the draft manuscript. All coauthors contributed to the writing of this article. K.T.K. is the guarantor. References [1] Health Statistics Quarterly. National statistics online. Available at. www. statistics.gov.uk. [2] Ware JE Jr, Kosinski M, Keller S. SF-36 physical and mental health summary scales: a user’s manual. Boston, MA: New England Medical Center, The Health Institute; 1994. [3] Weinstein MC, Stason WB. Foundations of cost-effectiveness analysis for health and medical practices. N Engl J Med 1977;296:716e21. [4] Tsuchiya A, Dolan P. The QALY model and individual preferences for health states and health profiles over time: a systematic review of the literature. Med Decis Making 2005;25:460e7. [5] About SF-6D. The University of Sheffield. Accessed March 2008. Available at: http://www.shef.ac.uk/scharr/sections/heds/mvh/sf-6d. [6] Brazier J, Roberts J, Deverill M. The estimation of a preference-based measure of health from the SF-36. J Health Econ 2002;21:271e92.

198

P.K. Myint et al. / Journal of Clinical Epidemiology 63 (2010) 192e198

[7] Brazier J, Usherwood T, Harper R, Thomas K. Deriving a preference-based single index from the UK SF-36 Health Survey. J Clin Epidemiol 1998;51:1115e28. [8] Lee BB, King MT, Simpson JM, Haran MJ, Stockler MR, Marial O, et al. Validity, responsiveness, and minimal important difference for the sf-6d health utility scale in a spinal cord injured population. Value Health 2008;. [Epub ahead of print]. [9] Khanna D, Furst DE, Wong WK, Tsevat J, Clements PJ, Park GS, et alScleroderma Collagen Type 1 Study Group. Reliability, validity, and minimally important differences of the SF-6D in systemic sclerosis. Qual Life Res 2007;16:1083e92. [10] Moock J, Kohlmann T. Comparing preference-based quality-of-life measures: results from rehabilitation patients with musculoskeletal, cardiovascular, or psychosomatic disorders. Qual Life Res 2008;17:485e95. [11] Marra CA, Marion SA, Guh DP, Najafzadeh M, Wolfe F, Esdaile JM, et al. Not all ‘‘quality-adjusted life years’’ are equal. J Clin Epidemiol 2007;60:616e24. [12] Myint PK, Luben RN, Surtees PG, Wainwright NWJ, Welch AA, Bingham SA, et al. Relation between self-reported physical functional health and chronic disease mortality in men and women in the European Prospective Investigation into Cancer (EPIC-Norfolk): a prospective population Study. Ann Epidemiol 2006;16:492e500. [13] Myint PK, Luben RN, Surtees PG, Wainwright NWJ, Welch AA, Bingham SA, et al. Self-reported mental health-related quality of life and mortality in men and women in the European Prospective Investigation into Cancer (EPIC-Norfolk): a prospective population study. Psychosom Med 2007;69:410e4. [14] Day N, Oakes S, Luben R, Khaw KT, Bingham S, Welch A. EPIC-Norfolk: study design and characteristics of the cohort. Br J Cancer 1999;80(Suppl 1):95e103. [15] Lohman T, Roche A, Martorell R. Anthropometric standardization reference manual. Champaign, IL: Human Kinetics Books; 1991. [16] Elias P, Halstead K, Prandy K. CASOC: Computer-assisted standard occupational coding. London: HMSO; 1993. [17] Shohaimi S, Luben R, Wareham N, Day N, Bingham S, Welch A, et al. Residential area deprivation predicts smoking habit independently of individual educational level and occupational social class. A cross sectional study in the Norfolk cohort of the European Investigation into Cancer (EPIC-Norfolk). J Epidemiol Community Health 2003;57:270e6. [18] Brazier JE, Harper R, Jones NMB, Ocathain A, Thomas KJ, Usherwood T, et al. Validating the SF-36 health survey questionnaire: new outcome measure for primary care. BMJ 1992;305:160e4. [19] Surtees PG, Wainwright NW, Khaw KT. Obesity, confidant support and functional health: cross-sectional evidence from the EPIC-Norfolk cohort. Int J Obes Relat Metab Disord 2004;28:748e58. [20] Myint PK. Healthy ageing: determinants and outcomes of self-reported functional health in the EPIC-Norfolk cohort. MD thesis, 2007.

[21] Idler EL, Russell LB, Davis D. Survival, functional limitations, and self-rated health in NHANES I epidemiologic follow-up study, 1992. Am J Epidemiol 2000;152:874e83. [22] Ferna´ndez M, Alarco´n GS, McGwin G Jr, Sanchez ML, Apte M, Vila´ LM, et al LUMINA Study Group. Using the Short Form 6D, as an overall measure of health, to predict damage accrual and mortality in patients with systemic lupus erythematosus: XLVII, results from a multiethnic US cohort. Arthritis Rheum 2007;57:986e92. [23] Khaw KT, Wareham N, Bingham S, Welch A, Luben R, Day N. Combined impact of health behaviours and mortality in men and women: the EPIC-Norfolk prospective population study. PLoS Med 2008;5(1):e12. [24] Myint PK, Luben RN, Surtees PG, Wainwright NWJ, Welch AA, Bingham SA, et al. Respiratory function and self-reported functional health: EPIC-Norfolk population study. Eur Respir J 2005;26: 494e502. [25] Myint PK, Welch AA, Luben RN, Wainwright NWJ, Surtees PG, Bingham SA, et al. Obesity indices and self-reported functional health in men and women in the EPIC-Norfolk. Obesity 2006;14: 884e93. [26] Myint PK, Welch AA, Bingham SA, Luben SR, Surtees PG, Wainwright NWJ, et al. Fruit and vegetable consumption and self-reported functional health in men and women in the European Prospective Investigation into Cancer-Norfolk (EPIC-Norfolk): a population based cross sectional study. Public Health Nutr 2007;10:34e41. [27] Myint PK, Surtees PG, Wainwright NWJ, Wareham NJ, Bingham SA, Luben RN, et al. Modifiable lifestyle behaviours and self-reported functional health in men and women in the European Prospective Investigation into Cancer (EPIC)eNorfolk: population based cross sectional study. Prev Med 2007;44:109e16. [28] Myint PK, Surtees PG, Wainwright NWJ, Luben RN, Welch AA, Bingham SA, et al. Physical health-related quality of life predicts stroke in the EPIC-Norfolk. Neurology 2007;69:2243e8. [29] Wee HL, Machin D, Loke WC, Li SC, Cheung YB, Luo N, et al. Assessing differences in utility scores: a comparison of four widely used preference-based instruments. Value Health 2007;10:256e65. [30] Hatoum HT, Brazier JE, Akhras KS. Comparison of the HUI3 with the SF-36 preference based SF-6D in a clinical trial setting. Value Health 2004;7:602e9. [31] Brazier J, Roberts J, Tsuchiya A, Busschbach J. A comparison of the EQ-5D and SF-6D across seven patient groups. Health Econ 2004;13: 873e84. [32] Fryback DG, Dunham NC, Palta M, Hanmer J, Buechner J, Cherepanov D, et al. US norms for six generic health-related quality-of-life indexes from the National Health Measurement study. Med Care 2007;45:1162e70.