- Email: [email protected]
Enhanced diabetes care to patients of south Asian ethnic origin (the United Kingdom Asian Diabetes Study): a cluster randomised controlled trial
Articles
Enhanced diabetes care to patients of south Asian ethnic origin (the United Kingdom Asian Diabetes Study): a cluster randomised controlled trial S Bellary*, J P O’Hare*, N T Raymond, A Gumber, S Mughal, A Szczepura, S Kumar, A H Barnett, for UKADS Study Group†
Summary Background Delivery of high-quality, evidence-based health care to deprived sectors of the community is a major goal for society. We investigated the effectiveness of a culturally sensitive, enhanced care package in UK general practices for improvement of cardiovascular risk factors in patients of south Asian origin with type 2 diabetes.
Lancet 2008; 371: 1769–76 See Editorial page 1723 See Comment page 1728 *Joint first authors
Methods In this cluster randomised controlled trial, 21 inner-city practices in the UK were assigned by simple randomisation to intervention (enhanced care including additional time with practice nurse and support from a link worker and diabetes-specialist nurse [nine practices; n=868]) or control (standard care [12 practices; n=618]) groups. All adult patients of south Asian origin with type 2 diabetes were eligible. Prescribing algorithms with clearly defined targets were provided for all practices. Primary outcomes were changes in blood pressure, total cholesterol, and glycaemic control (haemoglobin A1c) after 2 years. Analysis was by intention to treat. This trial is registered, number ISRCTN 38297969. Findings We recorded significant differences between treatment groups in diastolic blood pressure (1·91 [95% CI –2·88 to –0·94] mm Hg, p=0·0001) and mean arterial pressure (1·36 [–2·49 to –0·23] mm Hg, p=0·0180), after adjustment for confounders and clustering. We noted no significant differences between groups for total cholesterol (0·03 [–0·04 to 0·11] mmol/L), systolic blood pressure (–0·33 [–2·41 to 1·75] mm Hg), or HbA1c (–0·15% [–0·33 to 0·03]). Economic analysis suggests that the nurse-led intervention was not cost effective (incremental cost-effectiveness ratio £28 933 per QALY gained). Across the whole study population over the 2 years of the trial, systolic blood pressure, diastolic blood pressure, and cholesterol decreased significantly by 4·9 (95% CI 4·0–5·9) mm Hg, 3·8 (3·2–4·4) mm Hg, and 0·45 (0·40–0·51) mmol/L, respectively, and we recorded a small and non-significant increase for haemoglobin A1c (0·04% [–0·04 to 0·13]), p=0·290). Interpretation We recorded additional, although small, benefits from our culturally tailored care package that were greater than the secular changes achieved in the UK in recent years. Stricter targets in general practice and further measures to motivate patients are needed to achieve best possible health-care outcomes in south Asian patients with diabetes.
†Members listed at end of paper Heart of England NHS Foundation Trust, Birmingham, UK (S Bellary MRCP, S Mughal Dip Nursing, Prof A H Barnett FRCP); Warwick Medical School, Coventry, UK (J P O’Hare FRCP, N T Raymond MSc, A Gumber PhD, Prof A Szczepura DPhil, Prof S Kumar FRCP); and University of Birmingham, Birmingham, UK (S Bellary, Prof A H Barnett) Correspondence to: Prof Anthony H Barnett, Undergraduate Centre, Heart of England NHS Foundation Trust, Bordesley Green East, Birmingham B9 5SS, UK anthony.barnett@ heartofengland.nhs.uk
Funding Pfizer, Sanofi-Aventis, Servier Laboratories UK, Merck Sharp & Dohme/Schering-Plough, Takeda UK, Roche, Merck Pharma, Daiichi-Sankyo UK, Boehringer Ingelheim, Eli Lilly, Novo Nordisk, Bristol-Myers Squibb, Solvay Health Care, and Assurance Medical Society UK.
Introduction Patients of south Asian ethnic background (UK decennial census categories Indian, Pakistani, Bangladeshi, and other Asians) with type 2 diabetes present special management challenges.1,2 In the UK, prevalence of type 2 diabetes is four-fold to six-fold higher in people from south Asia than in white Europeans.3 Furthermore, onset can be more than a decade earlier and the risk of cardiovascular and renal complications greater in patients from south Asia, with higher morbidity and 50% higher mortality.4 Health-care delivery in this population is more challenging because of cultural, communication, and comprehension difficulties, which along with social deprivation further complicate the achievement of defined targets.5,6 Payments for UK general practices based on their achievement of quality (quality and outcomes framework [QOF])7 targets do not distinguish different ethnic groups. www.thelancet.com Vol 371 May 24, 2008
Enhanced care packages based in the community have been associated with improved metabolic outcomes in some ethnic groups8 but have not been fully assessed in large randomised controlled trials. Such trials are scarce in people of south Asian ethnic origin.9 The United Kingdom Asian Diabetes Study (UKADS) assessed a community-based complex intervention that aimed to reduce cardiovascular risk in south Asian people with type 2 diabetes. The intervention package was tailored to the needs of the south Asian community and consisted of additional time with a practice nurse, Asian link workers, and input from diabetes-specialist nurses, who were working to protocols to achieve clearly defined targets. The UKADS study hypothesis was that an enhanced care package for diabetes would improve cardiovascular risk profile in patients of south Asian origin, with established type 2 diabetes. 1769
Articles
Methods Study design and patients
See Online for webfigures 1–3
In line with recognised complex intervention evaluations10 and following a protocol informed by a pilot study,11 we undertook a large cluster randomised controlled trial from March, 2004 to April, 2007. 21 general practices (seven in Coventry [500 patients] and 14 in Birmingham, UK [986 patients]) with a very high proportion (more than 80%) of south Asian patients were included in this cluster randomised controlled trial. Between March 2004 and April 2005, nine practices were randomised to enhanced (intervention) and 12 to conventional (control) care. We used simple randomisation in both areas to achieve reasonable balance between groups. A common treatment algorithm was provided for control of blood pressure (webfigure 1), type 2 diabetes (webfigure 2), and lipid control (webfigure 3). All adult patients of south Asian origin with type 2 diabetes were eligible for inclusion in the study. There were no exclusion criteria.
Procedures Enhanced care included an additional practice nurse time (4 h per practice per week), supported by link workers and a community nurse specialising in diabetes. Patients in the intervention group were followed up on average every 2 months in clinics held every week by the practice nurses. Practice nurses had protected time to run a research diabetes clinic in intervention practices, and they worked with primary-care physicians to implement the protocol and encourage appropriate prescribing, provide face-to-face patient education in clinic setting, and achieve targets for blood pressure, lipid, and glycaemic control. Practice nurses were formally trained in diabetes and had 1:1 observed sessions with a diabetes-specialist nurse. All patients were contacted by a link worker before and between appointments to encourage clinic
Month 0: randomisation of practices
attendance. Additionally, link workers provided interpretation and additional educational input in local languages (Punjabi, Urdu, and Mirpuri) to patients in the community setting to improve compliance and understanding and to encourage dietary and lifestyle changes. Link workers attended research clinics in intervention practices. A total of five link workers were employed—three in Birmingham (14 practices) and two in Coventry (seven practices), with each one responsible for three or more practices. All link workers had attended a foundation course (equivalent to diploma) in diabetes management and care. The two community diabetes-specialist nurses covered the nine intervention practices and attended some research clinics every 6–8 weeks, providing additional educational and clinical support including insulin initiation, to the practice teams. Two specialist nurses were responsible for all 21 practices in the trial, one based in Coventry and one in Birmingham. All staff had formal training and experience in delivering diabetes care in the practice setting. The standard of care provided by the practice nurse and the link worker was monitored by the specialist nurse in observed sessions once every 3 months. General practitioners had overall responsibility for implementation of the study protocol within their practice and were involved in changing prescribing processes. Practices were encouraged to adhere to treatment protocols and to achieve targets. The study targets followed internationally accepted norms and were haemoglobin A1c of 7·0% (accepted target at the time of commencement of study), total cholesterol of 4·0 mmol/L, and blood pressure 130/80 mm Hg if no microvascular complications (as recommended by the Joint British Societies and international bodies)12–14 and 125/75 mm Hg if microalbuminuria or proteinuria was present. Control practices received the same treatment protocols, and practices managed patients with their existing resources.
Intervention
Control
General practitioner implemented study protocol. Clinical input from diabetes-specialist nurse
General practitioner implemented study protocol
Months 1–8: recruitment Patients recruited and baseline data collected by practice nurse, of patients and collection supported by link workers and diabetes-specialist nurse. EQ5D of baseline data questionnaire administered
Patients were recruited and baseline data collected, including EQ5D questionnaire, by practice and research nurses
Research clinic appointments every 2 months
Session (4 h) with practice nurse, supported by link workers and diabetes-specialist nurse. Guided by prescribing algorithm provided. Practice nurse consulted general practitioner when patients needed prescribing change
Routine practice nurse led diabetes clinics, but guided by prescribing algorithm. Practice nurse consulted general practitioner when patients needed prescribing change
Meetings for research monitoring (2–3 months apart)
Research team met with diabetes-specialist nurse and link worker to monitor recruitment and data collection, and to discuss and address issues of study conduct and management
Research team met with diabetes-specialist nurse and link worker to monitor recruitment and data collection, and to discuss and address issues of study conduct and management
Months 12–20: outcomes Outcomes assessed by practice nurse, link workers, and assessed diabetes-specialist nurse. Data from EQ5D questionnaire analysed
Outcomes assessed by practice nurse, link workers, and diabetes-specialist nurse. Data from EQ5D questionnaire analysed
EQ5D=EuroQol Five Dimensions.
Table 1: Components and timings of the complex intervention
1770
www.thelancet.com Vol 371 May 24, 2008
Articles
The study protocol was approved by East Birmingham and Coventry Primary Care Trust Ethics Committees. All patients provided written informed consent. Table 1 shows the components and timings of the complex intervention. Primary outcomes were follow-up measurements at 2 years for blood pressure, total cholesterol, and haemoglobin A1c, with secondary outcomes of waist circumference, body-mass index (BMI), Framingham 10 years coronary heart disease (CHD) risk score,15 microalbuminuria, and plasma creatinine. We passively monitored adverse events, and practices were encouraged to report any incidents related to the intervention.
3571 patients screened for eligibility (2070 in Birmingham and 1501 in Coventry) 1145 ineligible (non-south-Asian ethnic origin, type 1 diabetes, impaired glucose tolerance) 940 refused 1486 patients recruited from 21 general practices (7 in Coventry and 14 in Birmingham)
21 practices randomised
Statistical analysis Estimations of sample size were made on the basis of the observed differences and intraclass correlations ([ICC] defined as variance between groups/within groups) from the pilot study or an ICC=0·05, which is derived from published estimates for primary care studies.16,17 In all estimations, power was set to 80%, and the two-sided probability value to p=0·05. Estimates were made for differences in changes in systolic blood pressure (7 [SD 21·25] mm Hg, ICC=0·035), total cholesterol (0·45 [1·1] mmol/L, ICC=0·05), and haemoglobin A1c (0·75 [2·1], ICC=0·05). All estimates from these data values resulted in 16–18 clusters of 80–100 patients being needed, allowing for 10% drop-out rate. We selected these effect sizes since they were similar to those recorded in the pilot study, changes of this magnitude would be clinically significant, and they reflected prescribing algorithm targets. We analysed data with the SAS software package (version 9.1.3). We compared baseline variables between groups with χ2 tests of independence, with t tests for continuous variables, which were first assessed for normality. Primary and secondary outcomes were continuous. In the main intervention assessment, final measured outcomes were modelled, with grand mean-centred baseline measures included as covariates. To adjust for clustering and potential confounding effects, the SAS PROC MIXED procedure was used to fit hierarchical, combined fixed and random effects models.18,19 In all cases, mixed models included fixed effects for area (Birmingham vs Coventry), sex, age at diagnosis of diabetes, duration of diabetes, and corresponding grand mean-centred baseline measurement. For haemoglobin A1c, treatment with insulin at baseline was included in final models. We included terms for antihypertensive treatments, angiotensin-converting enzyme inhibitors (ACE), or angiotensin-receptor blockers (ARB) at baseline in models of blood pressure. For total cholesterol, statins and fibrates were included. Random effects were fitted, within a subject term for general practice, allowing for different intercepts and regression slopes for all individual practices (random coefficients models). www.thelancet.com Vol 371 May 24, 2008
868 patients to intervention (9 practices)
121 no follow-up data 46 lost 16 too ill to attend 35 refused 24 died
868 analysed by intention to treat
618 patients to control (12 practices)
87 no follow-up data 33 lost 4 too ill to attend 26 refused 24 died
618 analysed by intention to treat
Figure: Trial profile
We used restricted maximum likelihood (REML) models to analyse data. The correlation structure used in reported results was unstructured in all cases; variance components structures were considered. We used SAS graphics options to plot and assess residuals and influential data points, which were then removed and models re-run; results presented do not exclude outliers. For the main intention-to-treat analysis comparing outcomes, all patients were included. We analysed baseline and 2-year follow-up data measurements; for patients whose follow-up data were not available (figure), we inputted data with last observation carried forward (LOCF) method. Data were an interim value measured after 1 year, for around 50% of patients, or the baseline value. We undertook analyses with the same final models only for patients with complete data and only with those who had not died; although estimates of effect differed slightly, results and their interpretation were essentially the same. We gathered detailed data for staff salaries, travel and subsistence, equipment costs, payment to practices, and prescribing to estimate the net intervention cost over 2 years. Changes in quality-adjusted life years (QALY) between intervention and control groups were measured with the EQ5D (EuroQol Five Dimensions) questionnaire.20 This study is registered, number ISRCTN 38297969.
Role of the funding source The sponsors of the study had no role in study design, data collection, data analysis, data interpretation, or 1771
Articles
writing of the report. NTR, SB, JPO’H, AHB, AS, SK, and AG had full access to all the data in the study. All investigators and the UKADS Study Group had final responsibility for the decision to submit for publication.
Results The figure shows the trial profile. 1486 patients of south Asian ethnic origin, with established type 2 diabetes, consented to take part and were included in the study; 500 (34%) from Coventry and 986 (66%) from Birmingham. Table 2 shows the baseline risk-factor profile for the intervention and control groups. Mean age for the whole Intervention (N=868) Control (N=618)
Total
Sex Women
396 (46%)
313 (51%)
709 (48%)
Men
472 (54%)
304 (49%)
776 (52%)
Age (years) <45
131 (15%)
84 (14%)
215 (14%)
45–64
467 (54%)
363 (59%)
830 (56%)
≥65
270 (31%)
171 (28%)
441 (30%)
Duration of diabetes (years) 0–4
367 (42%)
222 (36%)
589 (40%)
5–9
230 (27%)
189 (31%)
419 (28%)
10–19
197 (23%)
161 (26%)
358 (24%)
72 (8%)
41 (7%)
113 (8%)
≥20 Treatment Insulin
161 (19%)
129 (21%)
290 (20%)
Oral
591 (68%)
429 (69%)
1020 (69%)
Diet only
116 (13%)
60 (10%)
176 (12%)
135 (16%)
86 (14%)
221 (15%)
59 (7%)
69 (11%)
128 (9%)
673 (78%)
462 (75%)
1135 (76%)
Smoking status* Current smoker Ex-smoker Non-smoker Weight (kg)
76·2 (14·6)
75·2 (14·6)
75·8 (14·5)
Waist (cm)
102·0 (11·5)
101·3 (12·3)
101·7 (11·8)
BMI (kg/m2)
28·5 (4·8)
28·6 (4·9)
28·5 (4·9)
Framingham 10 years CHD risk score
10·5 (8·8)
10·6 (8·8)
10·6 (8·8)
MAP (mm Hg)
101·7 (12·9)
102·9 (12·9)
102·2 (12·9)
Systolic BP (mm Hg)
139·4 (21·1)
141·1 (20·3)
140·1 (20·8)
Diastolic BP (mm Hg)
83·3 (11·0)
Risk factors profile
82·9 (11·0)
83·8 (11·1)
Total cholesterol (mmol/L)
4·7 (1·1)
4·7 (1·1)
4·7 (1·1)
Hb A1c (%)
8·2% (1·9)
8·2% (1·8)
8·2% (1·9)
Prescribed drugs All antihypertensive drugs
475 (55%)
342 (55%)
817 (55%)
ACE/ARB
321 (37%)
246 (40%)
567 (38%)
Statins*
438 (50%)
273 (44%)
711 (48%)
Data are number (%) or mean (SD). Data are missing for duration of diabetes (n=7), smoking status (2), weight (2), waist circumference (5), BMI (10), total cholesterol (2), and Hb A1c (13). BMI=body-mass index. MAP=mean arterial pressure. BP=blood pressure. CHD=coronary heart disease. ACE=angiotensin-converting enzyme. ARB=angiotensinreceptor blocker. *Significance difference between groups at 5% level: smoking status χ²=9·01, p=0·01; statins χ²=5·72, p=0·02.
Table 2: Baseline characteristics
1772
group was 57·0 (SD 11·9) years. Differences observed between groups for sex, age, duration of diabetes, and treatment for diabetes were not significant. The proportion of current smokers was much the same in both groups, but more patients in the control group than in the intervention group were ex-smokers. We recorded no differences in weight, BMI, or waist circumference measurements. More intervention than control patients were treated with statins. At baseline, 268 (18%) patients (150 [17%] in the intervention group and 118 [19%] in the control group) had evidence of existing coronary heart disease or previous cardiovascular events, angina, myocardial infarction, cardiovascular accident, coronary artery bypass graft, or other heart problems. At baseline we measured urinary albumin to creatinine ratio for 1389 (93%) patients (807 [93%] in the intervention and 582 [94%] in the control group) and noted that microalbuminuria (defined as a ratio >2·5 in men and >3·5 in women) was present in 268 (19%) patients (161 [20%] vs 107 [18%]). We detected significant proteinuria, defined as albumin to creatinine ratio of more than 25·0, in 114 (8%) patients (61 [8%] vs 53 [9%]). The prevalence of combined microalbuminuria or proteinuria was 28%, with no difference between intervention and control groups (222 [28%] vs 160 [27%]). With the Framingham equation, mean 10-year risk score for coronary heart disease was 10·6 (SD 8·8), with no difference between treatment groups (table 2). During 2 years of follow-up, 48 (3%) patients died— 24 (3%) in the intervention group and 24 (4%) in the control group. New cardiovascular events were recorded for 97 (7%) patients—62 (7%) in the intervention group and 35 (6%) in the control group. None of these small differences between intervention and control groups was significant. In a post-hoc analyses, patients with coronary heart disease at baseline were more likely to die (18 [7%] vs 30 [2%]) or to have events of coronary heart disease during follow-up (34 [13%] vs 63 [5%]) than were those with no evidence of this disease, irrespective of treatment group. We recorded no adverse events related to intervention during the study period. Table 3 shows the results for comparison of outcomes between intervention and control groups. After 2 years we recorded a reduction of 5·1 (from 139·4 to 134·3) mm Hg in systolic blood pressure and 4·5 (from 82·9 to 78·4) mm Hg in diastolic blood pressure for patients in the intervention group compared with 4·7 (from 141·1 to 136·4) mm Hg and 2·9 (from 83·8 to 81·0) mm Hg, respectively, in the control group. Results from t tests showed significant differences in favour of the intervention group for diastolic blood pressure (p=0·0065) and haemoglobin A1c (p=0·0371) (table 3). After adjustment for potential confounders, we noted significant advantages for the intervention group for diastolic blood pressure (p<0·0001) and mean arterial pressure (p=0·0003) (table 3). In final models taking www.thelancet.com Vol 371 May 24, 2008
Articles
Difference between means (95% CI)* Differences least squares means (95% CI)† Differences least squares means (95% CI)‡ Primary outcomes MAP (mm Hg)
–1·2 (–2·5 to 0·1) (p=0·061)
–2·0 ( –3·1 to –0·9) (p=0·0003)
–1·4 (–2·5 to –0·2) (p=0·018)
Systolic BP (mm Hg)
–0·4 (–2·3 to 1·5) (p=0·66)
–1·4 (–3·0 to 0·2) (p=0·082)
–0·3 (–2·4 to 1·8) (p=0·76)
Diastolic BP (mm Hg)
–1·6 (–2·8 to –0·5) (p=0·007)
–2·3 (–3·3 to –1·3) (p<0·0001)
–1·91 (–2·9 to –0·9) (p=0·0001)
Total cholesterol (mmol/L) Hb A1c (%)
0·01 (–0·11 to 0·12) (p=0·88)
0·02 (–0·07 to 0·12) (p=0·64)
0·03 (–0·04 to 0·11) (p=0·37)
–0·18 (–0·34 to –0·01) (p=0·037)
–0·13 (–0·28 to 0·02) (p=0·080)
–0·15 (–0·33 to 0·03) (p=0·11)
0·06 (–0·56 to 0·68) (p=0·85)
–0·08 (–0·62 to 0·46) (p=0·77)
Secondary outcomes CHD risk (Framingham)§ Waist (cm) BMI (kg/m²)
–0·3 (–1·0 to 0·5) (p=0·51) 0·38 (0·20 to 0·55) (p<0·0001)
–0·1 (–0·8 to 0·6) (p=0·86) 0·40 (0·22 to 0·57) (p<0·0001)
0·01 (–0·57 to 0·59) (p=0·97) –0·2 (–1·3 to 0·9) (p=0·67) 0·40 (0·20 to 0·60) (p<0·0001)
BMI=body-mass index. MAP=mean arterial pressure. BP=blood pressure. CHD=coronary heart disease. *Crude differences based on t-test comparison, no adjustment. †Differences based on fixed effects model, adjusted for confounding. ‡Differences based on mixed model, adjusted for confounding and clustering. §Framingham risk of coronary heart disease estimated only for patients aged 30–74 years at baseline (n=1376).
Table 3: Differences in outcomes between intervention and control group, adjusted for potential confounding and clustering
clustering effects into account, significant effects persisted for the intervention group for both mean arterial pressure (p=0·018) and diastolic blood pressure (p=0·0001) (table 3). BMI was significantly increased in the intervention group (p<0·0001; table 3). Other differences in primary and secondary outcomes were small and not significant after adjustment for confounding and clustering (table 3). The number of patients with microalbuminuria or proteinuria increased from 382 (28%) at baseline to 474 (32%) after 2 years, with no significant difference between the intervention and control groups. Patients at high renal risk, defined by plasma creatinine greater than 120 µmol/L for women and greater than 150 µmol/L for men, increased from 61 (4%) at baseline to 83 (6%) after 2 years, with no difference between treatment groups. When we combined all patients from both groups after 2 years, we noted an overall decrease of 4·9 (95% CI 4·0–5·9) mm Hg in systolic blood pressure (p<0·0001), 3·8 (3·2–4·4) mm Hg in diastolic blood pressure (p<0·0001), and 4·2 (3·6–4·8) mm Hg in mean arterial pressure (p<0·0001). Total cholesterol decreased by 0·45 (0·40–0·51) mmol/L (p<0·0001). We recorded a small and non-significant increase for haemoglobin A1c (0·04% [–0·04 to 0·13], p=0·290). After 2 years of follow-up in post-hoc analyses, the number of patients given antihypertensive drugs had increased to 1119 (75%) overall, with no difference between groups (660 [76%] in the intervention and 459 [74%] in the control group). Treatment with statins had increased, with 540 (64%) patients being treated in the intervention group compared with 389 (65%) in the control group. The use of ACE inhibitors or ARBs increased substantially from 321 (37%) to 569 (66%) in the intervention and from 246 (40%) to 389 (62%) in the control group; the groups did not differ significantly. A similar proportion of patients were treated with insulin at baseline (table 2). After 2 years, more patients in the intervention group than in the control group had www.thelancet.com Vol 371 May 24, 2008
Cost (GB£) Staff salaries* Payment to practices†
224 774 50 000
Travel and subsistence
17 720
Clinical equipment
11 060
Total enhanced diabetes care service over 2 years Per patient enhanced service cost over 2 years
303 554 406
Prescribing cost Per patient net prescribing cost for non-diabetic drugs over 2 years
16
Per patient net prescribing cost for diabetic drugs over 2 years
12
Per patient net prescribing cost over 2 years Total per patient incremental net cost of intervention over 2 years Incremental cost per QALY gained‡
28 434 28 933
All costs are incremental cost between intervention and control based on the actual expenditure incurred on different items during the study. QALY=quality-adjusted life year. *Staff salaries covered clinical time of two specialist nurses and five link workers. The salaries included national insurance and pension contributions. †The net amount paid to nine intervention practices to implement the intervention over 2 years. ‡Per patient quality-adjusted life year gain over 2 years was 0·015.
Table 4: Intervention costs and incremental cost-effectiveness over 2 years
started insulin therapy (47 [8%] vs 23 [5%]), but this finding was not significant (relative risk 1·44 [95% CI 0·89–2·34]). In post-hoc analyses, the number of patients achieving the study targets for blood pressure was 310 (36%) of 868 in the intervention group versus 191 (31%) of 618 in the control group, for cholesterol 411 (47%) of 867 versus 311 (50%) of 617, and for haemoglobin A1c 275 (32%) of 858 versus 165 (27%) of 615. For the QOF targets, the corresponding number for blood pressure of less than 145/85 mm Hg was 575 (66%) versus 346 (56%), for cholesterol less than 5 mmol/L 700 (81%) versus 509 (83%), and for haemoglobin A1c less than 7·5% 377 (44%) versus 240 (39%). 1773
Articles
Table 4 provides a detailed cost breakdown for the intervention. Over 2 years, the cost of intervention per patient was GB£434 (£406 net service and £28 net prescribing costs). Despite patients’ overall quality of life deteriorating over 2 years, the resultant net change in quality of life in the intervention group compared with control group was positive, although small (0·015). Thus, we calculated the incremental cost-effectiveness ratio to be £28 933 per QALY gained.
Discussion Our results confirm that the achievement of targets set by national and international advisory bodies poses a major challenge for south Asian ethnic groups in inner-city general practices.12–14,21 At baseline, many of our patients had haemoglobin A1c greater than 7%, blood pressure greater than 130/80 mm Hg, and total cholesterol greater than 4 mmol/L, which are higher than targets recommended by international standards for diabetes care. After 2 years in which secular changes included the pay for performance initiative, we noted significant improvements in blood pressure and total cholesterol across the whole study population, but no change in haemoglobin A1c. A reduction in blood pressure has been associated with rapid reduction in cardiovascular risk in many studies.22–24 The relation between blood pressure and cardiovascular risk is such that a sustained reduction of 5 mm Hg would confer substantial protection from cardiovascular events.25 The improvements in blood pressure and cholesterol that we recorded were associated with increased prescribing of antihypertensive agents and statins, and are consistent with improvements reported by several other investigators after the introduction of QOF initiatives.26 The mortality noted during our study, together with the frequency of cardiovascular events at baseline and followup, confirm that the south Asian group which we investigated has a high cardiovascular risk and that substantial benefits could be obtained by aggressive reduction in risk factors. The failure to prevent the increase in microalbuminuria despite a 5 mm Hg decrease in blood pressure is surprising and suggests that lower targets could be needed for this group. A comparison of intervention and control groups after 2 years showed significant differences for diastolic blood pressure and mean arterial pressure after adjustment for confounding and clustering. Although systolic blood pressure was lower in the intervention than in the control group, the result was not significant. The reductions seen in diastolic blood pressure were comparable with those recorded in our pilot study,11 but the reduction in systolic blood pressure was less than was previously achieved. The fairly young age of onset and ethnic origin might be a factor in this finding, and a more pronounced diastolic effect has been reported in some other studies.27 We noted a small but significant increase in BMI in the intervention group. This finding could be because of the 1774
increased use of insulin in the intervention group, but other factors such as poor adherence to lifestyle advice might have contributed. The absence of significant improvement in haemoglobin A1c could be partly due to the natural disease progression that is commonly seen in type 2 diabetes;28 haemoglobin A1c tended to rise in the control group, whereas it remained stable over 2 years in the intervention group. In view of the health-care resources provided, we find it disappointing that neither the QOF incentives nor our culturally sensitive enhanced care package significantly effected glycaemic control. Despite clear evidence of failure to reach target levels of haemoglobin A1c via diet and oral antidiabetic therapy, we recorded only a small increase in the percentage of patients given insulin in both groups. Even though the intervention was supported by nurses specialised in diabetes, who have experience of insulin initiation and patient education, this support seems to have had only a small effect in terms of behavioural change or patient acceptance of insulin. Initiation of insulin in many primary-care practices in the UK is fairly new, and building up confidence of both the health-care team and south Asian patients might be as important as any financial incentives that the health-care team receive. Changes in patient behaviour through motivation and patient education might take longer than the 2-year follow-up in this study. Alternative methods of motivation, including structured patient education29 and more aggressive insulin initiation, might be needed. Significant improvements in performance indicators were noted across the UK general practice after the introduction of the QOF initiatives, and even the control practices in our study probably benefited from these changes. Many fewer patients achieved the study targets than did those meeting the QOF targets, suggesting adherence to treatment protocols was poor in both groups. Despite additional nursing resources, we recorded small improvements in the intervention group, which might have been due to reluctance of health professionals to intensify treatments and achieve tighter targets beyond those already set in the QOF initiative. Such factors will not be exclusive to south Asian patients receiving primary care in the UK and might apply to other racial groups and health-care settings. The economic analysis shows that the financial investment needed over 2 years did not produce sufficient health-related gain in quality of life to make such a nurse-led intervention clearly cost effective. At £28 933 per QALY gained, compared with an indicative norm of £30 000 per QALY,30 wide-scale implementation is not suggested without improvement in effectiveness. In our study, not all patients were receiving statins at the start of the trial. The proportion of patients receiving statins increased over 2 years and was comparable to www.thelancet.com Vol 371 May 24, 2008
Articles
other published reports in patients with diabetes.31 However, only two-thirds of patients were receiving statins despite our protocol stating that all patients should be prescribed them. The precise reason for this finding is unclear. One reason could be that target cholesterol concentrations in the QOF were 5 mmol/L. Therefore patients whose cholesterol concentrations achieved these targets did not receive statins. We used the LOCF method in our analyses, which we acknowledge has its weaknesses. However, an analysis of complete data only, produced very similar results. A further limitation of our study is the inability to assess the relative contributions of individual components of the intervention; such challenges are inherent to the assessment of complex interventions. Despite the modest clinical outcomes achieved in the study, evidence suggests that intensive management can improve outcomes in type 2 diabetes. Although the intervention might need development, the strength of the study is that it applies rigorous scientific evaluation to a socially deprived ethnic minority group. Substantial difficulties in recruiting and retaining individuals of south Asian ethnic origin have been reported previously by several investigators,32,33 which might account for the scarcity of large-scale studies in this population.34 However, our experience suggests that recruitment and retention is possible in this hardto-reach group. Our results suggest that small but sustained improvements in blood pressure can be achieved through the introduction of a culturally sensitive, enhanced care package for south Asian patients in addition to improvements from the QOF financial incentives. Improvement in glycaemic control remains a major challenge, and further work to enhance effectiveness of health-care delivery in general practice and to improve motivation is clearly needed for this group if health-care inequalities are to be reduced. Although progress has been made, a substantial challenge remains to achieve the more stringent targets that are recommended by national and international expert advisory bodies. Contributors JPO’H and SK, and AHB had the original idea for and designed the project. NTR contributed to study design and conducted/supervised statistical analysis. SB and SM were responsible for day-to-day running of the project, helping with data collection and analysis. AG undertook the analysis of economic data. AS helped in data analysis and interpretation. All authors contributed to writing the paper and data interpretation. UKADS Writing Group S Bellary, J P O’Hare, N T Raymond, S Mughal, A Gumber, A Szczepura, S Kumar, A H Barnett. UKADS Study Group University of Birmingham and Heart of England NHS Trust, Birmingham, UK: S Bellary, A H Barnett, S Mughal, S Begum, T Kauser, N Mirza, A N Dixon, W M Hanif, A Rahim, W Malik, A Jones, R A Bhatti, M M Alvi, A Akthar, R A S Sangra, N H Bangash, M D Sheik, A G Hakeem, B G Najak, S A Latif, J S Sanghera, A U Shah, T Sen-Gupta, G O’Gara, P S Moonga, S H Khattak, P Machin, F Hartland, H Kaur,
www.thelancet.com Vol 371 May 24, 2008
N Janood, J Dhalival, T Fatima, J Taylor, D Fitzpatrick, J Lucas, S Hemming; and Warwick Medical School, Warwick University and University Hospital Coventry and Warwickshire, Coventry, UK: J P O’Hare, N T Raymond, S Kumar, A Szczepura, K Johal, A Gumber, I C Agarwal, D K Mistry, F F Lyall, M R Dhadhania, K L Kakad, U Jetty, J F Sihota, S Mall, K Webb, S Khatoon, R Parker, P Claire, G Turner. Conflict of interest statement JPO’H, SK, and AHB have received research grants and lecture fees from the companies that sponsored this study. All other authors declare that they have no conflict of interest. Acknowledgments We thank Nigel Stallard, Professor of Medical Statistics, Warwick Medical School, Warwick, UK, for comments and discussion relating to SAS PROC MIXED, hierarchical models, and adjustment for clustering. This study was funded by financial support in the form of grants for the UKAD Study from Pfizer, Sanofi-Aventis, Servier Laboratories UK, Merck Sharp & Dohme/Schering-Plough, Takeda UK, Roche, Merck Pharma, Daiichi-Sankyo UK, Boehringer Ingelheim, Eli Lilly, NovoNordisk, Bristol-Myers Squibb, Solvay Health Care, and Assurance Medical Society, UK. References 1 Greenhalgh PM. Diabetes in British south Asians: nature, nurture, and culture. Diabet Med 1997; 14: 10–18. 2 Barnett AH, Dixon AN, Bellary S, et al. Type 2 diabetes and cardiovascular risk in the UK south Asian community. Diabetologia 2006; 49: 2234–46. 3 Mather HM, Keen H. The Southall Diabetes Survey: prevalence of known diabetes in Asians and Europeans. Br Med J (Clin Res Ed) 1985; 291: 1081–84. 4 Chaturvedi N, Fuller JH. Ethnic differences in mortality from cardiovascular disease in the UK: do they persist in people with diabetes? J Epidemiol Community Health 1996; 50: 137–39. 5 Health Survey for England 2004: health of ethnic minorities— full report. http://www.ic.nhs.uk/statistics (accessed Feb 21, 2008). 6 Stone M, Pound E, Pancholi A, Farooqi A, Khunti K. Empowering patients with diabetes: a qualitative primary care study focusing on South Asians in Leicester, UK. Fam Pract 2005; 22: 647–52. 7 Shekelle P. New contract for general practitioners. BMJ 2003; 326: 457–58. 8 Brown SA, Garcia AA, Kouzekanani K, Hanis CL. Culturally competent diabetes self-management education for Mexican Americans: the Starr County border health initiative. Diabetes Care 2002; 25: 259–68. 9 Gammon BD, Gunarathne A. It’s time to reappraise recruitment of South Asians to clinical trials. BMJ 2008; 336: 46. 10 Campbell M, Fitzpatrick R, Haines A, et al. Framework for design and evaluation of complex interventions to improve health. BMJ 2000; 321: 694–96. 11 O’Hare JP, Raymond NT, Mughal S, et al. Evaluation of delivery of enhanced diabetes care to patients of South Asian ethnicity: the United Kingdom Asian Diabetes Study (UKADS). Diabet Med 2004; 21: 1357–65. 12 Position statement. American Diabetic Association. Diabetes Care 2007; 30 (suppl 1): S4–41. 13 Chobanian AV, Bakris GL, Black HR, et al. The seventh report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure: the JNC 7 report. JAMA 2003; 289: 2560–72. 14 Grundy SM, Cleeman JI, Merz CN, et al. Implications of recent clinical trials for the National Cholesterol Education Program Adult Treatment Panel III guidelines. Circulation 2004; 110: 227–39. 15 Anderson KM, Wilson PW, Odell PM, Kannel WB. An updated coronary risk profile. A statement for health professionals. Circulation 1991; 83: 356–62. 16 Smeeth L, Ng ES. Intraclass correlation coefficients for cluster randomized trials in primary care: data from the MRC Trial of the Assessment and Management of Older People in the Community. Control Clin Trials 2002; 23: 409–21. 17 Underwood M, Barnett A, Hajioff S. Cluster randomization: a trap for the unwary. Br J Gen Pract 1998; 48: 1089–90.
1775
Articles
18
19
20
21
22
23
24
25
26
1776
Singer JD. Using SAS PROC MIXED to fit multilevel models, hierarchical models, and individual growth models. J Educ Behav Stat 1998; 24: 323–55. Sullivan LM, Dukes KA, Lofina E. Tutorial in biostatistics: an introduction to hierarchical linear modelling. Stat Med 1999; 18: 855–88. Dolan PG, Kind P, Williams A. A social tariff for EuroQol: results from a UK general population survey. York: Centre for Health Economics, University of York, 1995. Williams B, Poulter NR, Brown MJ, et al. Guidelines for management of hypertension: report of the fourth working party of the British Hypertension Society, 2004-BHS IV. J Hum Hypertens 2004; 18: 139–85. Collins R, MacMahon S. Blood pressure, antihypertensive drug treatment and the risks of stroke and of coronary heart disease. Br Med Bull 1994; 50: 272–98. Gueyffier F, Boutitie F, Boissel JP, et al. Effect of antihypertensive drug treatment on cardiovascular outcomes in women and men. A meta-analysis of individual patient data from randomized, controlled trials. The INDANA Investigators. Ann Intern Med 1997; 126: 761–67. Hansson L, Zanchetti A, Carruthers SG, et al, for the HOT Study Group. Effects of intensive blood-pressure lowering and low-dose aspirin in patients with hypertension: principal results of the Hypertension Optimal Treatment (HOT) randomised trial. Lancet 1998; 351: 1755–62. Prospective studies collaboration. Age-specific relevance of usual blood pressure to vascular mortality: a meta-analysis of individual data for one million adults in 61 prospective studies. Lancet 2002; 360: 1903–13. Khunti K, Gadsby R, Millett C, Majeed A, Davies M. Quality of diabetes care in the UK: comparison of published quality-of-care reports with results of the Quality and Outcomes Framework for Diabetes. Diabet Med 2007; 24: 1436–41.
27
28
29
30 31
32 33
34
Heart Outcomes Prevention Evaluation Study Investigators. Effects of ramipril on cardiovascular and microvascular outcomes in people with diabetes mellitus: results of the HOPE study and MICRO-HOPE substudy. Lancet 2000; 355: 253–59. Wright A, Burden AC, Paisey RB, Cull CA, Holman RR. Sulfonylurea inadequacy: efficacy of addition of insulin over 6 years in patients with type 2 diabetes in the U.K. Prospective Diabetes Study (UKPDS 57). Diabetes Care 2002; 25: 330–36. Davies MJ, Heller S, Skinner TC, et al. Effectiveness of the diabetes education and self management for ongoing and newly diagnosed (DESMOND) programme for people with newly diagnosed type 2 diabetes: cluster randomised controlled trial. BMJ 2008; 336: 491–95. Raftery J. NICE: faster access to modern treatments? Analysis of guidance on health technologies. BMJ 2001; 323: 1300–03. Millet C, Gray J, Saxena S, Netuveli G, Khunti K, Majeed A. Ethnic disparities in Diabetes Management and Pay-for-Performance in the UK: the Wandsworth Prospective Diabetes Study. PLos Med 2007; 4: 1087–92. Hunt S, Bhopal R. Self reports in research with non-English speakers. BMJ 2003; 327: 352–53. Hussain-Gambles M, Leese B, Atkin K, Brown J, Mason S, Tovey P. Involving South Asian patients in clinical trials. Health Technol Assess 2004; 8: 1–109. Saxena S, Misra T, Car J, Netuveli G, Smith R, Majeed A. Systematic review of primary healthcare interventions to improve diabetes outcomes in minority ethnic groups. J Ambul Care Manage 2007; 30: 218–30.
www.thelancet.com Vol 371 May 24, 2008
Enhanced diabetes care to patients of south Asian ethnic origin (the United Kingdom Asian Diabetes Study): a cluster randomised controlled trial S Bellary*, J P O’Hare*, N T Raymond, A Gumber, S Mughal, A Szczepura, S Kumar, A H Barnett, for UKADS Study Group†
Summary Background Delivery of high-quality, evidence-based health care to deprived sectors of the community is a major goal for society. We investigated the effectiveness of a culturally sensitive, enhanced care package in UK general practices for improvement of cardiovascular risk factors in patients of south Asian origin with type 2 diabetes.
Lancet 2008; 371: 1769–76 See Editorial page 1723 See Comment page 1728 *Joint first authors
Methods In this cluster randomised controlled trial, 21 inner-city practices in the UK were assigned by simple randomisation to intervention (enhanced care including additional time with practice nurse and support from a link worker and diabetes-specialist nurse [nine practices; n=868]) or control (standard care [12 practices; n=618]) groups. All adult patients of south Asian origin with type 2 diabetes were eligible. Prescribing algorithms with clearly defined targets were provided for all practices. Primary outcomes were changes in blood pressure, total cholesterol, and glycaemic control (haemoglobin A1c) after 2 years. Analysis was by intention to treat. This trial is registered, number ISRCTN 38297969. Findings We recorded significant differences between treatment groups in diastolic blood pressure (1·91 [95% CI –2·88 to –0·94] mm Hg, p=0·0001) and mean arterial pressure (1·36 [–2·49 to –0·23] mm Hg, p=0·0180), after adjustment for confounders and clustering. We noted no significant differences between groups for total cholesterol (0·03 [–0·04 to 0·11] mmol/L), systolic blood pressure (–0·33 [–2·41 to 1·75] mm Hg), or HbA1c (–0·15% [–0·33 to 0·03]). Economic analysis suggests that the nurse-led intervention was not cost effective (incremental cost-effectiveness ratio £28 933 per QALY gained). Across the whole study population over the 2 years of the trial, systolic blood pressure, diastolic blood pressure, and cholesterol decreased significantly by 4·9 (95% CI 4·0–5·9) mm Hg, 3·8 (3·2–4·4) mm Hg, and 0·45 (0·40–0·51) mmol/L, respectively, and we recorded a small and non-significant increase for haemoglobin A1c (0·04% [–0·04 to 0·13]), p=0·290). Interpretation We recorded additional, although small, benefits from our culturally tailored care package that were greater than the secular changes achieved in the UK in recent years. Stricter targets in general practice and further measures to motivate patients are needed to achieve best possible health-care outcomes in south Asian patients with diabetes.
†Members listed at end of paper Heart of England NHS Foundation Trust, Birmingham, UK (S Bellary MRCP, S Mughal Dip Nursing, Prof A H Barnett FRCP); Warwick Medical School, Coventry, UK (J P O’Hare FRCP, N T Raymond MSc, A Gumber PhD, Prof A Szczepura DPhil, Prof S Kumar FRCP); and University of Birmingham, Birmingham, UK (S Bellary, Prof A H Barnett) Correspondence to: Prof Anthony H Barnett, Undergraduate Centre, Heart of England NHS Foundation Trust, Bordesley Green East, Birmingham B9 5SS, UK anthony.barnett@ heartofengland.nhs.uk
Funding Pfizer, Sanofi-Aventis, Servier Laboratories UK, Merck Sharp & Dohme/Schering-Plough, Takeda UK, Roche, Merck Pharma, Daiichi-Sankyo UK, Boehringer Ingelheim, Eli Lilly, Novo Nordisk, Bristol-Myers Squibb, Solvay Health Care, and Assurance Medical Society UK.
Introduction Patients of south Asian ethnic background (UK decennial census categories Indian, Pakistani, Bangladeshi, and other Asians) with type 2 diabetes present special management challenges.1,2 In the UK, prevalence of type 2 diabetes is four-fold to six-fold higher in people from south Asia than in white Europeans.3 Furthermore, onset can be more than a decade earlier and the risk of cardiovascular and renal complications greater in patients from south Asia, with higher morbidity and 50% higher mortality.4 Health-care delivery in this population is more challenging because of cultural, communication, and comprehension difficulties, which along with social deprivation further complicate the achievement of defined targets.5,6 Payments for UK general practices based on their achievement of quality (quality and outcomes framework [QOF])7 targets do not distinguish different ethnic groups. www.thelancet.com Vol 371 May 24, 2008
Enhanced care packages based in the community have been associated with improved metabolic outcomes in some ethnic groups8 but have not been fully assessed in large randomised controlled trials. Such trials are scarce in people of south Asian ethnic origin.9 The United Kingdom Asian Diabetes Study (UKADS) assessed a community-based complex intervention that aimed to reduce cardiovascular risk in south Asian people with type 2 diabetes. The intervention package was tailored to the needs of the south Asian community and consisted of additional time with a practice nurse, Asian link workers, and input from diabetes-specialist nurses, who were working to protocols to achieve clearly defined targets. The UKADS study hypothesis was that an enhanced care package for diabetes would improve cardiovascular risk profile in patients of south Asian origin, with established type 2 diabetes. 1769
Articles
Methods Study design and patients
See Online for webfigures 1–3
In line with recognised complex intervention evaluations10 and following a protocol informed by a pilot study,11 we undertook a large cluster randomised controlled trial from March, 2004 to April, 2007. 21 general practices (seven in Coventry [500 patients] and 14 in Birmingham, UK [986 patients]) with a very high proportion (more than 80%) of south Asian patients were included in this cluster randomised controlled trial. Between March 2004 and April 2005, nine practices were randomised to enhanced (intervention) and 12 to conventional (control) care. We used simple randomisation in both areas to achieve reasonable balance between groups. A common treatment algorithm was provided for control of blood pressure (webfigure 1), type 2 diabetes (webfigure 2), and lipid control (webfigure 3). All adult patients of south Asian origin with type 2 diabetes were eligible for inclusion in the study. There were no exclusion criteria.
Procedures Enhanced care included an additional practice nurse time (4 h per practice per week), supported by link workers and a community nurse specialising in diabetes. Patients in the intervention group were followed up on average every 2 months in clinics held every week by the practice nurses. Practice nurses had protected time to run a research diabetes clinic in intervention practices, and they worked with primary-care physicians to implement the protocol and encourage appropriate prescribing, provide face-to-face patient education in clinic setting, and achieve targets for blood pressure, lipid, and glycaemic control. Practice nurses were formally trained in diabetes and had 1:1 observed sessions with a diabetes-specialist nurse. All patients were contacted by a link worker before and between appointments to encourage clinic
Month 0: randomisation of practices
attendance. Additionally, link workers provided interpretation and additional educational input in local languages (Punjabi, Urdu, and Mirpuri) to patients in the community setting to improve compliance and understanding and to encourage dietary and lifestyle changes. Link workers attended research clinics in intervention practices. A total of five link workers were employed—three in Birmingham (14 practices) and two in Coventry (seven practices), with each one responsible for three or more practices. All link workers had attended a foundation course (equivalent to diploma) in diabetes management and care. The two community diabetes-specialist nurses covered the nine intervention practices and attended some research clinics every 6–8 weeks, providing additional educational and clinical support including insulin initiation, to the practice teams. Two specialist nurses were responsible for all 21 practices in the trial, one based in Coventry and one in Birmingham. All staff had formal training and experience in delivering diabetes care in the practice setting. The standard of care provided by the practice nurse and the link worker was monitored by the specialist nurse in observed sessions once every 3 months. General practitioners had overall responsibility for implementation of the study protocol within their practice and were involved in changing prescribing processes. Practices were encouraged to adhere to treatment protocols and to achieve targets. The study targets followed internationally accepted norms and were haemoglobin A1c of 7·0% (accepted target at the time of commencement of study), total cholesterol of 4·0 mmol/L, and blood pressure 130/80 mm Hg if no microvascular complications (as recommended by the Joint British Societies and international bodies)12–14 and 125/75 mm Hg if microalbuminuria or proteinuria was present. Control practices received the same treatment protocols, and practices managed patients with their existing resources.
Intervention
Control
General practitioner implemented study protocol. Clinical input from diabetes-specialist nurse
General practitioner implemented study protocol
Months 1–8: recruitment Patients recruited and baseline data collected by practice nurse, of patients and collection supported by link workers and diabetes-specialist nurse. EQ5D of baseline data questionnaire administered
Patients were recruited and baseline data collected, including EQ5D questionnaire, by practice and research nurses
Research clinic appointments every 2 months
Session (4 h) with practice nurse, supported by link workers and diabetes-specialist nurse. Guided by prescribing algorithm provided. Practice nurse consulted general practitioner when patients needed prescribing change
Routine practice nurse led diabetes clinics, but guided by prescribing algorithm. Practice nurse consulted general practitioner when patients needed prescribing change
Meetings for research monitoring (2–3 months apart)
Research team met with diabetes-specialist nurse and link worker to monitor recruitment and data collection, and to discuss and address issues of study conduct and management
Research team met with diabetes-specialist nurse and link worker to monitor recruitment and data collection, and to discuss and address issues of study conduct and management
Months 12–20: outcomes Outcomes assessed by practice nurse, link workers, and assessed diabetes-specialist nurse. Data from EQ5D questionnaire analysed
Outcomes assessed by practice nurse, link workers, and diabetes-specialist nurse. Data from EQ5D questionnaire analysed
EQ5D=EuroQol Five Dimensions.
Table 1: Components and timings of the complex intervention
1770
www.thelancet.com Vol 371 May 24, 2008
Articles
The study protocol was approved by East Birmingham and Coventry Primary Care Trust Ethics Committees. All patients provided written informed consent. Table 1 shows the components and timings of the complex intervention. Primary outcomes were follow-up measurements at 2 years for blood pressure, total cholesterol, and haemoglobin A1c, with secondary outcomes of waist circumference, body-mass index (BMI), Framingham 10 years coronary heart disease (CHD) risk score,15 microalbuminuria, and plasma creatinine. We passively monitored adverse events, and practices were encouraged to report any incidents related to the intervention.
3571 patients screened for eligibility (2070 in Birmingham and 1501 in Coventry) 1145 ineligible (non-south-Asian ethnic origin, type 1 diabetes, impaired glucose tolerance) 940 refused 1486 patients recruited from 21 general practices (7 in Coventry and 14 in Birmingham)
21 practices randomised
Statistical analysis Estimations of sample size were made on the basis of the observed differences and intraclass correlations ([ICC] defined as variance between groups/within groups) from the pilot study or an ICC=0·05, which is derived from published estimates for primary care studies.16,17 In all estimations, power was set to 80%, and the two-sided probability value to p=0·05. Estimates were made for differences in changes in systolic blood pressure (7 [SD 21·25] mm Hg, ICC=0·035), total cholesterol (0·45 [1·1] mmol/L, ICC=0·05), and haemoglobin A1c (0·75 [2·1], ICC=0·05). All estimates from these data values resulted in 16–18 clusters of 80–100 patients being needed, allowing for 10% drop-out rate. We selected these effect sizes since they were similar to those recorded in the pilot study, changes of this magnitude would be clinically significant, and they reflected prescribing algorithm targets. We analysed data with the SAS software package (version 9.1.3). We compared baseline variables between groups with χ2 tests of independence, with t tests for continuous variables, which were first assessed for normality. Primary and secondary outcomes were continuous. In the main intervention assessment, final measured outcomes were modelled, with grand mean-centred baseline measures included as covariates. To adjust for clustering and potential confounding effects, the SAS PROC MIXED procedure was used to fit hierarchical, combined fixed and random effects models.18,19 In all cases, mixed models included fixed effects for area (Birmingham vs Coventry), sex, age at diagnosis of diabetes, duration of diabetes, and corresponding grand mean-centred baseline measurement. For haemoglobin A1c, treatment with insulin at baseline was included in final models. We included terms for antihypertensive treatments, angiotensin-converting enzyme inhibitors (ACE), or angiotensin-receptor blockers (ARB) at baseline in models of blood pressure. For total cholesterol, statins and fibrates were included. Random effects were fitted, within a subject term for general practice, allowing for different intercepts and regression slopes for all individual practices (random coefficients models). www.thelancet.com Vol 371 May 24, 2008
868 patients to intervention (9 practices)
121 no follow-up data 46 lost 16 too ill to attend 35 refused 24 died
868 analysed by intention to treat
618 patients to control (12 practices)
87 no follow-up data 33 lost 4 too ill to attend 26 refused 24 died
618 analysed by intention to treat
Figure: Trial profile
We used restricted maximum likelihood (REML) models to analyse data. The correlation structure used in reported results was unstructured in all cases; variance components structures were considered. We used SAS graphics options to plot and assess residuals and influential data points, which were then removed and models re-run; results presented do not exclude outliers. For the main intention-to-treat analysis comparing outcomes, all patients were included. We analysed baseline and 2-year follow-up data measurements; for patients whose follow-up data were not available (figure), we inputted data with last observation carried forward (LOCF) method. Data were an interim value measured after 1 year, for around 50% of patients, or the baseline value. We undertook analyses with the same final models only for patients with complete data and only with those who had not died; although estimates of effect differed slightly, results and their interpretation were essentially the same. We gathered detailed data for staff salaries, travel and subsistence, equipment costs, payment to practices, and prescribing to estimate the net intervention cost over 2 years. Changes in quality-adjusted life years (QALY) between intervention and control groups were measured with the EQ5D (EuroQol Five Dimensions) questionnaire.20 This study is registered, number ISRCTN 38297969.
Role of the funding source The sponsors of the study had no role in study design, data collection, data analysis, data interpretation, or 1771
Articles
writing of the report. NTR, SB, JPO’H, AHB, AS, SK, and AG had full access to all the data in the study. All investigators and the UKADS Study Group had final responsibility for the decision to submit for publication.
Results The figure shows the trial profile. 1486 patients of south Asian ethnic origin, with established type 2 diabetes, consented to take part and were included in the study; 500 (34%) from Coventry and 986 (66%) from Birmingham. Table 2 shows the baseline risk-factor profile for the intervention and control groups. Mean age for the whole Intervention (N=868) Control (N=618)
Total
Sex Women
396 (46%)
313 (51%)
709 (48%)
Men
472 (54%)
304 (49%)
776 (52%)
Age (years) <45
131 (15%)
84 (14%)
215 (14%)
45–64
467 (54%)
363 (59%)
830 (56%)
≥65
270 (31%)
171 (28%)
441 (30%)
Duration of diabetes (years) 0–4
367 (42%)
222 (36%)
589 (40%)
5–9
230 (27%)
189 (31%)
419 (28%)
10–19
197 (23%)
161 (26%)
358 (24%)
72 (8%)
41 (7%)
113 (8%)
≥20 Treatment Insulin
161 (19%)
129 (21%)
290 (20%)
Oral
591 (68%)
429 (69%)
1020 (69%)
Diet only
116 (13%)
60 (10%)
176 (12%)
135 (16%)
86 (14%)
221 (15%)
59 (7%)
69 (11%)
128 (9%)
673 (78%)
462 (75%)
1135 (76%)
Smoking status* Current smoker Ex-smoker Non-smoker Weight (kg)
76·2 (14·6)
75·2 (14·6)
75·8 (14·5)
Waist (cm)
102·0 (11·5)
101·3 (12·3)
101·7 (11·8)
BMI (kg/m2)
28·5 (4·8)
28·6 (4·9)
28·5 (4·9)
Framingham 10 years CHD risk score
10·5 (8·8)
10·6 (8·8)
10·6 (8·8)
MAP (mm Hg)
101·7 (12·9)
102·9 (12·9)
102·2 (12·9)
Systolic BP (mm Hg)
139·4 (21·1)
141·1 (20·3)
140·1 (20·8)
Diastolic BP (mm Hg)
83·3 (11·0)
Risk factors profile
82·9 (11·0)
83·8 (11·1)
Total cholesterol (mmol/L)
4·7 (1·1)
4·7 (1·1)
4·7 (1·1)
Hb A1c (%)
8·2% (1·9)
8·2% (1·8)
8·2% (1·9)
Prescribed drugs All antihypertensive drugs
475 (55%)
342 (55%)
817 (55%)
ACE/ARB
321 (37%)
246 (40%)
567 (38%)
Statins*
438 (50%)
273 (44%)
711 (48%)
Data are number (%) or mean (SD). Data are missing for duration of diabetes (n=7), smoking status (2), weight (2), waist circumference (5), BMI (10), total cholesterol (2), and Hb A1c (13). BMI=body-mass index. MAP=mean arterial pressure. BP=blood pressure. CHD=coronary heart disease. ACE=angiotensin-converting enzyme. ARB=angiotensinreceptor blocker. *Significance difference between groups at 5% level: smoking status χ²=9·01, p=0·01; statins χ²=5·72, p=0·02.
Table 2: Baseline characteristics
1772
group was 57·0 (SD 11·9) years. Differences observed between groups for sex, age, duration of diabetes, and treatment for diabetes were not significant. The proportion of current smokers was much the same in both groups, but more patients in the control group than in the intervention group were ex-smokers. We recorded no differences in weight, BMI, or waist circumference measurements. More intervention than control patients were treated with statins. At baseline, 268 (18%) patients (150 [17%] in the intervention group and 118 [19%] in the control group) had evidence of existing coronary heart disease or previous cardiovascular events, angina, myocardial infarction, cardiovascular accident, coronary artery bypass graft, or other heart problems. At baseline we measured urinary albumin to creatinine ratio for 1389 (93%) patients (807 [93%] in the intervention and 582 [94%] in the control group) and noted that microalbuminuria (defined as a ratio >2·5 in men and >3·5 in women) was present in 268 (19%) patients (161 [20%] vs 107 [18%]). We detected significant proteinuria, defined as albumin to creatinine ratio of more than 25·0, in 114 (8%) patients (61 [8%] vs 53 [9%]). The prevalence of combined microalbuminuria or proteinuria was 28%, with no difference between intervention and control groups (222 [28%] vs 160 [27%]). With the Framingham equation, mean 10-year risk score for coronary heart disease was 10·6 (SD 8·8), with no difference between treatment groups (table 2). During 2 years of follow-up, 48 (3%) patients died— 24 (3%) in the intervention group and 24 (4%) in the control group. New cardiovascular events were recorded for 97 (7%) patients—62 (7%) in the intervention group and 35 (6%) in the control group. None of these small differences between intervention and control groups was significant. In a post-hoc analyses, patients with coronary heart disease at baseline were more likely to die (18 [7%] vs 30 [2%]) or to have events of coronary heart disease during follow-up (34 [13%] vs 63 [5%]) than were those with no evidence of this disease, irrespective of treatment group. We recorded no adverse events related to intervention during the study period. Table 3 shows the results for comparison of outcomes between intervention and control groups. After 2 years we recorded a reduction of 5·1 (from 139·4 to 134·3) mm Hg in systolic blood pressure and 4·5 (from 82·9 to 78·4) mm Hg in diastolic blood pressure for patients in the intervention group compared with 4·7 (from 141·1 to 136·4) mm Hg and 2·9 (from 83·8 to 81·0) mm Hg, respectively, in the control group. Results from t tests showed significant differences in favour of the intervention group for diastolic blood pressure (p=0·0065) and haemoglobin A1c (p=0·0371) (table 3). After adjustment for potential confounders, we noted significant advantages for the intervention group for diastolic blood pressure (p<0·0001) and mean arterial pressure (p=0·0003) (table 3). In final models taking www.thelancet.com Vol 371 May 24, 2008
Articles
Difference between means (95% CI)* Differences least squares means (95% CI)† Differences least squares means (95% CI)‡ Primary outcomes MAP (mm Hg)
–1·2 (–2·5 to 0·1) (p=0·061)
–2·0 ( –3·1 to –0·9) (p=0·0003)
–1·4 (–2·5 to –0·2) (p=0·018)
Systolic BP (mm Hg)
–0·4 (–2·3 to 1·5) (p=0·66)
–1·4 (–3·0 to 0·2) (p=0·082)
–0·3 (–2·4 to 1·8) (p=0·76)
Diastolic BP (mm Hg)
–1·6 (–2·8 to –0·5) (p=0·007)
–2·3 (–3·3 to –1·3) (p<0·0001)
–1·91 (–2·9 to –0·9) (p=0·0001)
Total cholesterol (mmol/L) Hb A1c (%)
0·01 (–0·11 to 0·12) (p=0·88)
0·02 (–0·07 to 0·12) (p=0·64)
0·03 (–0·04 to 0·11) (p=0·37)
–0·18 (–0·34 to –0·01) (p=0·037)
–0·13 (–0·28 to 0·02) (p=0·080)
–0·15 (–0·33 to 0·03) (p=0·11)
0·06 (–0·56 to 0·68) (p=0·85)
–0·08 (–0·62 to 0·46) (p=0·77)
Secondary outcomes CHD risk (Framingham)§ Waist (cm) BMI (kg/m²)
–0·3 (–1·0 to 0·5) (p=0·51) 0·38 (0·20 to 0·55) (p<0·0001)
–0·1 (–0·8 to 0·6) (p=0·86) 0·40 (0·22 to 0·57) (p<0·0001)
0·01 (–0·57 to 0·59) (p=0·97) –0·2 (–1·3 to 0·9) (p=0·67) 0·40 (0·20 to 0·60) (p<0·0001)
BMI=body-mass index. MAP=mean arterial pressure. BP=blood pressure. CHD=coronary heart disease. *Crude differences based on t-test comparison, no adjustment. †Differences based on fixed effects model, adjusted for confounding. ‡Differences based on mixed model, adjusted for confounding and clustering. §Framingham risk of coronary heart disease estimated only for patients aged 30–74 years at baseline (n=1376).
Table 3: Differences in outcomes between intervention and control group, adjusted for potential confounding and clustering
clustering effects into account, significant effects persisted for the intervention group for both mean arterial pressure (p=0·018) and diastolic blood pressure (p=0·0001) (table 3). BMI was significantly increased in the intervention group (p<0·0001; table 3). Other differences in primary and secondary outcomes were small and not significant after adjustment for confounding and clustering (table 3). The number of patients with microalbuminuria or proteinuria increased from 382 (28%) at baseline to 474 (32%) after 2 years, with no significant difference between the intervention and control groups. Patients at high renal risk, defined by plasma creatinine greater than 120 µmol/L for women and greater than 150 µmol/L for men, increased from 61 (4%) at baseline to 83 (6%) after 2 years, with no difference between treatment groups. When we combined all patients from both groups after 2 years, we noted an overall decrease of 4·9 (95% CI 4·0–5·9) mm Hg in systolic blood pressure (p<0·0001), 3·8 (3·2–4·4) mm Hg in diastolic blood pressure (p<0·0001), and 4·2 (3·6–4·8) mm Hg in mean arterial pressure (p<0·0001). Total cholesterol decreased by 0·45 (0·40–0·51) mmol/L (p<0·0001). We recorded a small and non-significant increase for haemoglobin A1c (0·04% [–0·04 to 0·13], p=0·290). After 2 years of follow-up in post-hoc analyses, the number of patients given antihypertensive drugs had increased to 1119 (75%) overall, with no difference between groups (660 [76%] in the intervention and 459 [74%] in the control group). Treatment with statins had increased, with 540 (64%) patients being treated in the intervention group compared with 389 (65%) in the control group. The use of ACE inhibitors or ARBs increased substantially from 321 (37%) to 569 (66%) in the intervention and from 246 (40%) to 389 (62%) in the control group; the groups did not differ significantly. A similar proportion of patients were treated with insulin at baseline (table 2). After 2 years, more patients in the intervention group than in the control group had www.thelancet.com Vol 371 May 24, 2008
Cost (GB£) Staff salaries* Payment to practices†
224 774 50 000
Travel and subsistence
17 720
Clinical equipment
11 060
Total enhanced diabetes care service over 2 years Per patient enhanced service cost over 2 years
303 554 406
Prescribing cost Per patient net prescribing cost for non-diabetic drugs over 2 years
16
Per patient net prescribing cost for diabetic drugs over 2 years
12
Per patient net prescribing cost over 2 years Total per patient incremental net cost of intervention over 2 years Incremental cost per QALY gained‡
28 434 28 933
All costs are incremental cost between intervention and control based on the actual expenditure incurred on different items during the study. QALY=quality-adjusted life year. *Staff salaries covered clinical time of two specialist nurses and five link workers. The salaries included national insurance and pension contributions. †The net amount paid to nine intervention practices to implement the intervention over 2 years. ‡Per patient quality-adjusted life year gain over 2 years was 0·015.
Table 4: Intervention costs and incremental cost-effectiveness over 2 years
started insulin therapy (47 [8%] vs 23 [5%]), but this finding was not significant (relative risk 1·44 [95% CI 0·89–2·34]). In post-hoc analyses, the number of patients achieving the study targets for blood pressure was 310 (36%) of 868 in the intervention group versus 191 (31%) of 618 in the control group, for cholesterol 411 (47%) of 867 versus 311 (50%) of 617, and for haemoglobin A1c 275 (32%) of 858 versus 165 (27%) of 615. For the QOF targets, the corresponding number for blood pressure of less than 145/85 mm Hg was 575 (66%) versus 346 (56%), for cholesterol less than 5 mmol/L 700 (81%) versus 509 (83%), and for haemoglobin A1c less than 7·5% 377 (44%) versus 240 (39%). 1773
Articles
Table 4 provides a detailed cost breakdown for the intervention. Over 2 years, the cost of intervention per patient was GB£434 (£406 net service and £28 net prescribing costs). Despite patients’ overall quality of life deteriorating over 2 years, the resultant net change in quality of life in the intervention group compared with control group was positive, although small (0·015). Thus, we calculated the incremental cost-effectiveness ratio to be £28 933 per QALY gained.
Discussion Our results confirm that the achievement of targets set by national and international advisory bodies poses a major challenge for south Asian ethnic groups in inner-city general practices.12–14,21 At baseline, many of our patients had haemoglobin A1c greater than 7%, blood pressure greater than 130/80 mm Hg, and total cholesterol greater than 4 mmol/L, which are higher than targets recommended by international standards for diabetes care. After 2 years in which secular changes included the pay for performance initiative, we noted significant improvements in blood pressure and total cholesterol across the whole study population, but no change in haemoglobin A1c. A reduction in blood pressure has been associated with rapid reduction in cardiovascular risk in many studies.22–24 The relation between blood pressure and cardiovascular risk is such that a sustained reduction of 5 mm Hg would confer substantial protection from cardiovascular events.25 The improvements in blood pressure and cholesterol that we recorded were associated with increased prescribing of antihypertensive agents and statins, and are consistent with improvements reported by several other investigators after the introduction of QOF initiatives.26 The mortality noted during our study, together with the frequency of cardiovascular events at baseline and followup, confirm that the south Asian group which we investigated has a high cardiovascular risk and that substantial benefits could be obtained by aggressive reduction in risk factors. The failure to prevent the increase in microalbuminuria despite a 5 mm Hg decrease in blood pressure is surprising and suggests that lower targets could be needed for this group. A comparison of intervention and control groups after 2 years showed significant differences for diastolic blood pressure and mean arterial pressure after adjustment for confounding and clustering. Although systolic blood pressure was lower in the intervention than in the control group, the result was not significant. The reductions seen in diastolic blood pressure were comparable with those recorded in our pilot study,11 but the reduction in systolic blood pressure was less than was previously achieved. The fairly young age of onset and ethnic origin might be a factor in this finding, and a more pronounced diastolic effect has been reported in some other studies.27 We noted a small but significant increase in BMI in the intervention group. This finding could be because of the 1774
increased use of insulin in the intervention group, but other factors such as poor adherence to lifestyle advice might have contributed. The absence of significant improvement in haemoglobin A1c could be partly due to the natural disease progression that is commonly seen in type 2 diabetes;28 haemoglobin A1c tended to rise in the control group, whereas it remained stable over 2 years in the intervention group. In view of the health-care resources provided, we find it disappointing that neither the QOF incentives nor our culturally sensitive enhanced care package significantly effected glycaemic control. Despite clear evidence of failure to reach target levels of haemoglobin A1c via diet and oral antidiabetic therapy, we recorded only a small increase in the percentage of patients given insulin in both groups. Even though the intervention was supported by nurses specialised in diabetes, who have experience of insulin initiation and patient education, this support seems to have had only a small effect in terms of behavioural change or patient acceptance of insulin. Initiation of insulin in many primary-care practices in the UK is fairly new, and building up confidence of both the health-care team and south Asian patients might be as important as any financial incentives that the health-care team receive. Changes in patient behaviour through motivation and patient education might take longer than the 2-year follow-up in this study. Alternative methods of motivation, including structured patient education29 and more aggressive insulin initiation, might be needed. Significant improvements in performance indicators were noted across the UK general practice after the introduction of the QOF initiatives, and even the control practices in our study probably benefited from these changes. Many fewer patients achieved the study targets than did those meeting the QOF targets, suggesting adherence to treatment protocols was poor in both groups. Despite additional nursing resources, we recorded small improvements in the intervention group, which might have been due to reluctance of health professionals to intensify treatments and achieve tighter targets beyond those already set in the QOF initiative. Such factors will not be exclusive to south Asian patients receiving primary care in the UK and might apply to other racial groups and health-care settings. The economic analysis shows that the financial investment needed over 2 years did not produce sufficient health-related gain in quality of life to make such a nurse-led intervention clearly cost effective. At £28 933 per QALY gained, compared with an indicative norm of £30 000 per QALY,30 wide-scale implementation is not suggested without improvement in effectiveness. In our study, not all patients were receiving statins at the start of the trial. The proportion of patients receiving statins increased over 2 years and was comparable to www.thelancet.com Vol 371 May 24, 2008
Articles
other published reports in patients with diabetes.31 However, only two-thirds of patients were receiving statins despite our protocol stating that all patients should be prescribed them. The precise reason for this finding is unclear. One reason could be that target cholesterol concentrations in the QOF were 5 mmol/L. Therefore patients whose cholesterol concentrations achieved these targets did not receive statins. We used the LOCF method in our analyses, which we acknowledge has its weaknesses. However, an analysis of complete data only, produced very similar results. A further limitation of our study is the inability to assess the relative contributions of individual components of the intervention; such challenges are inherent to the assessment of complex interventions. Despite the modest clinical outcomes achieved in the study, evidence suggests that intensive management can improve outcomes in type 2 diabetes. Although the intervention might need development, the strength of the study is that it applies rigorous scientific evaluation to a socially deprived ethnic minority group. Substantial difficulties in recruiting and retaining individuals of south Asian ethnic origin have been reported previously by several investigators,32,33 which might account for the scarcity of large-scale studies in this population.34 However, our experience suggests that recruitment and retention is possible in this hardto-reach group. Our results suggest that small but sustained improvements in blood pressure can be achieved through the introduction of a culturally sensitive, enhanced care package for south Asian patients in addition to improvements from the QOF financial incentives. Improvement in glycaemic control remains a major challenge, and further work to enhance effectiveness of health-care delivery in general practice and to improve motivation is clearly needed for this group if health-care inequalities are to be reduced. Although progress has been made, a substantial challenge remains to achieve the more stringent targets that are recommended by national and international expert advisory bodies. Contributors JPO’H and SK, and AHB had the original idea for and designed the project. NTR contributed to study design and conducted/supervised statistical analysis. SB and SM were responsible for day-to-day running of the project, helping with data collection and analysis. AG undertook the analysis of economic data. AS helped in data analysis and interpretation. All authors contributed to writing the paper and data interpretation. UKADS Writing Group S Bellary, J P O’Hare, N T Raymond, S Mughal, A Gumber, A Szczepura, S Kumar, A H Barnett. UKADS Study Group University of Birmingham and Heart of England NHS Trust, Birmingham, UK: S Bellary, A H Barnett, S Mughal, S Begum, T Kauser, N Mirza, A N Dixon, W M Hanif, A Rahim, W Malik, A Jones, R A Bhatti, M M Alvi, A Akthar, R A S Sangra, N H Bangash, M D Sheik, A G Hakeem, B G Najak, S A Latif, J S Sanghera, A U Shah, T Sen-Gupta, G O’Gara, P S Moonga, S H Khattak, P Machin, F Hartland, H Kaur,
www.thelancet.com Vol 371 May 24, 2008
N Janood, J Dhalival, T Fatima, J Taylor, D Fitzpatrick, J Lucas, S Hemming; and Warwick Medical School, Warwick University and University Hospital Coventry and Warwickshire, Coventry, UK: J P O’Hare, N T Raymond, S Kumar, A Szczepura, K Johal, A Gumber, I C Agarwal, D K Mistry, F F Lyall, M R Dhadhania, K L Kakad, U Jetty, J F Sihota, S Mall, K Webb, S Khatoon, R Parker, P Claire, G Turner. Conflict of interest statement JPO’H, SK, and AHB have received research grants and lecture fees from the companies that sponsored this study. All other authors declare that they have no conflict of interest. Acknowledgments We thank Nigel Stallard, Professor of Medical Statistics, Warwick Medical School, Warwick, UK, for comments and discussion relating to SAS PROC MIXED, hierarchical models, and adjustment for clustering. This study was funded by financial support in the form of grants for the UKAD Study from Pfizer, Sanofi-Aventis, Servier Laboratories UK, Merck Sharp & Dohme/Schering-Plough, Takeda UK, Roche, Merck Pharma, Daiichi-Sankyo UK, Boehringer Ingelheim, Eli Lilly, NovoNordisk, Bristol-Myers Squibb, Solvay Health Care, and Assurance Medical Society, UK. References 1 Greenhalgh PM. Diabetes in British south Asians: nature, nurture, and culture. Diabet Med 1997; 14: 10–18. 2 Barnett AH, Dixon AN, Bellary S, et al. Type 2 diabetes and cardiovascular risk in the UK south Asian community. Diabetologia 2006; 49: 2234–46. 3 Mather HM, Keen H. The Southall Diabetes Survey: prevalence of known diabetes in Asians and Europeans. Br Med J (Clin Res Ed) 1985; 291: 1081–84. 4 Chaturvedi N, Fuller JH. Ethnic differences in mortality from cardiovascular disease in the UK: do they persist in people with diabetes? J Epidemiol Community Health 1996; 50: 137–39. 5 Health Survey for England 2004: health of ethnic minorities— full report. http://www.ic.nhs.uk/statistics (accessed Feb 21, 2008). 6 Stone M, Pound E, Pancholi A, Farooqi A, Khunti K. Empowering patients with diabetes: a qualitative primary care study focusing on South Asians in Leicester, UK. Fam Pract 2005; 22: 647–52. 7 Shekelle P. New contract for general practitioners. BMJ 2003; 326: 457–58. 8 Brown SA, Garcia AA, Kouzekanani K, Hanis CL. Culturally competent diabetes self-management education for Mexican Americans: the Starr County border health initiative. Diabetes Care 2002; 25: 259–68. 9 Gammon BD, Gunarathne A. It’s time to reappraise recruitment of South Asians to clinical trials. BMJ 2008; 336: 46. 10 Campbell M, Fitzpatrick R, Haines A, et al. Framework for design and evaluation of complex interventions to improve health. BMJ 2000; 321: 694–96. 11 O’Hare JP, Raymond NT, Mughal S, et al. Evaluation of delivery of enhanced diabetes care to patients of South Asian ethnicity: the United Kingdom Asian Diabetes Study (UKADS). Diabet Med 2004; 21: 1357–65. 12 Position statement. American Diabetic Association. Diabetes Care 2007; 30 (suppl 1): S4–41. 13 Chobanian AV, Bakris GL, Black HR, et al. The seventh report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure: the JNC 7 report. JAMA 2003; 289: 2560–72. 14 Grundy SM, Cleeman JI, Merz CN, et al. Implications of recent clinical trials for the National Cholesterol Education Program Adult Treatment Panel III guidelines. Circulation 2004; 110: 227–39. 15 Anderson KM, Wilson PW, Odell PM, Kannel WB. An updated coronary risk profile. A statement for health professionals. Circulation 1991; 83: 356–62. 16 Smeeth L, Ng ES. Intraclass correlation coefficients for cluster randomized trials in primary care: data from the MRC Trial of the Assessment and Management of Older People in the Community. Control Clin Trials 2002; 23: 409–21. 17 Underwood M, Barnett A, Hajioff S. Cluster randomization: a trap for the unwary. Br J Gen Pract 1998; 48: 1089–90.
1775
Articles
18
19
20
21
22
23
24
25
26
1776
Singer JD. Using SAS PROC MIXED to fit multilevel models, hierarchical models, and individual growth models. J Educ Behav Stat 1998; 24: 323–55. Sullivan LM, Dukes KA, Lofina E. Tutorial in biostatistics: an introduction to hierarchical linear modelling. Stat Med 1999; 18: 855–88. Dolan PG, Kind P, Williams A. A social tariff for EuroQol: results from a UK general population survey. York: Centre for Health Economics, University of York, 1995. Williams B, Poulter NR, Brown MJ, et al. Guidelines for management of hypertension: report of the fourth working party of the British Hypertension Society, 2004-BHS IV. J Hum Hypertens 2004; 18: 139–85. Collins R, MacMahon S. Blood pressure, antihypertensive drug treatment and the risks of stroke and of coronary heart disease. Br Med Bull 1994; 50: 272–98. Gueyffier F, Boutitie F, Boissel JP, et al. Effect of antihypertensive drug treatment on cardiovascular outcomes in women and men. A meta-analysis of individual patient data from randomized, controlled trials. The INDANA Investigators. Ann Intern Med 1997; 126: 761–67. Hansson L, Zanchetti A, Carruthers SG, et al, for the HOT Study Group. Effects of intensive blood-pressure lowering and low-dose aspirin in patients with hypertension: principal results of the Hypertension Optimal Treatment (HOT) randomised trial. Lancet 1998; 351: 1755–62. Prospective studies collaboration. Age-specific relevance of usual blood pressure to vascular mortality: a meta-analysis of individual data for one million adults in 61 prospective studies. Lancet 2002; 360: 1903–13. Khunti K, Gadsby R, Millett C, Majeed A, Davies M. Quality of diabetes care in the UK: comparison of published quality-of-care reports with results of the Quality and Outcomes Framework for Diabetes. Diabet Med 2007; 24: 1436–41.
27
28
29
30 31
32 33
34
Heart Outcomes Prevention Evaluation Study Investigators. Effects of ramipril on cardiovascular and microvascular outcomes in people with diabetes mellitus: results of the HOPE study and MICRO-HOPE substudy. Lancet 2000; 355: 253–59. Wright A, Burden AC, Paisey RB, Cull CA, Holman RR. Sulfonylurea inadequacy: efficacy of addition of insulin over 6 years in patients with type 2 diabetes in the U.K. Prospective Diabetes Study (UKPDS 57). Diabetes Care 2002; 25: 330–36. Davies MJ, Heller S, Skinner TC, et al. Effectiveness of the diabetes education and self management for ongoing and newly diagnosed (DESMOND) programme for people with newly diagnosed type 2 diabetes: cluster randomised controlled trial. BMJ 2008; 336: 491–95. Raftery J. NICE: faster access to modern treatments? Analysis of guidance on health technologies. BMJ 2001; 323: 1300–03. Millet C, Gray J, Saxena S, Netuveli G, Khunti K, Majeed A. Ethnic disparities in Diabetes Management and Pay-for-Performance in the UK: the Wandsworth Prospective Diabetes Study. PLos Med 2007; 4: 1087–92. Hunt S, Bhopal R. Self reports in research with non-English speakers. BMJ 2003; 327: 352–53. Hussain-Gambles M, Leese B, Atkin K, Brown J, Mason S, Tovey P. Involving South Asian patients in clinical trials. Health Technol Assess 2004; 8: 1–109. Saxena S, Misra T, Car J, Netuveli G, Smith R, Majeed A. Systematic review of primary healthcare interventions to improve diabetes outcomes in minority ethnic groups. J Ambul Care Manage 2007; 30: 218–30.
www.thelancet.com Vol 371 May 24, 2008