OUP user menu

Obesity and kidney disease in type 1 and 2 diabetes: an analysis of the National Diabetes Audit

C.J. Hill, C.R. Cardwell, A.P. Maxwell, R.J. Young, B. Matthews, D.J. O’Donoghue, D.G. Fogarty
DOI: http://dx.doi.org/10.1093/qjmed/hct123 933-942 First published online: 21 May 2013

Abstract

Background: Obesity is increasingly prevalent in many countries. Obesity is a major risk factor for the development of type 2 diabetes but its relationship with diabetic kidney disease (DKD) remains unclear. Some studies have suggested that the metabolic syndrome (including obesity) may be associated with DKD in type 1 diabetes.

Aim: To investigate the association between obesity and DKD.

Design: Retrospective cross-sectional study.

Methods: National Diabetes Audit data were available for the 2007–08 cycle. Type 1 and 2 diabetes patients with both a valid serum creatinine and urinary albumin:creatinine ratio were included. DKD was defined as an estimated glomerular filtration rate (eGFR) <60 ml/min/1.73 m2, albuminuria or both. Logistic regression was used to analyse associations of obesity (body mass index ≥30 kg/m2) and other variables including year of birth, year of diagnosis, ethnicity and stage of kidney disease.

Results: A total of 58 791 type 1 and 733 769 type 2 diabetes patients were included in the analysis. After adjustment, when compared with type 1 diabetes patients with normal renal function those with DKD were up to twice as likely to be obese. Type 2 DKD patients were also more likely to be obese. For example, type 2 diabetes patients with an eGFR <15 ml/min/1.73 m2 and normoalbuminuria, microalbuminuria or macroalbuminuria were all more likely to be obese; odds ratios (95% CI) 1.65 (1.3–2.1), 1.56 (1.28–1.92) and 1.27 (1.05–1.54), respectively.

Conclusions: This study has highlighted a strong association between obesity and kidney disease in type 1 diabetes and confirmed their association in type 2 diabetes.

Background

Obesity is increasingly common in many countries.1 It is a major contributor to the development of insulin resistance which is a feature of both the metabolic syndrome and type 2 diabetes mellitus.2,3 However, the relationship between obesity and diabetic kidney disease (DKD) remains unclear.

Obesity has been associated with the development of kidney disease in the absence of diabetes.4 Some obese patients develop glomerular hyperfiltration which can progress to proteinuria, often in the nephrotic range, followed by a progressive reduction in glomerular filtration rate (GFR).5 Biopsy features in such patients have shown a mix of lesions from early glomerulomegaly to advanced focal segmental glomerulosclerosis.6 However, the role of obesity in the development and progression of kidney disease remains a subject of debate.7 Some studies have suggested that, after adjustment for other cardiovascular risk factors, obesity is not associated with progression of chronic kidney disease (CKD).8 Nevertheless, increased physical exercise and weight loss have been associated with a reduction in the risk of development and progression of proteinuria in non-diabetic obese patients with CKD.9

The majority of DKD patients have type 2 diabetes, which is associated with obesity.1 However, obesity may also be associated with DKD in type 1 diabetes. Recent evidence suggests that familial history of type 2 diabetes may be associated with cardiovascular risk and risk of DKD in type 1 diabetes.10–12 This could relate to inherited hyperglycaemia-induced epigenetic modifications resulting in predisposition to insulin resistance, obesity and nephropathy.13

Other demographic and clinical factors may affect the risk of obesity as well as the risk of developing diabetes. For example, lower socioeconomic status has been associated with a higher risk of obesity and glucose intolerance as well as poorer treatment concordance.14,15

We aimed to investigate the prevalence of, and factors associated with, obesity in DKD patients.

Subjects and methods

The National Diabetes Audit (NDA) in England and Wales is delivered by the NHS Information Centre, working in conjunction with Diabetes UK and Diabetes Health Intelligence (Yorkshire and Humber Public Health Observatory). It is responsible for collating and analysing data on four key areas of diabetes management (registration, care processes, treatment targets and complications) and providing feedback to participating practices.16

The NDA receives submissions from primary and secondary care facilities in England and Wales. Each audit cycle lasts from 1 January to 31 March of the subsequent year. Data are collated from healthcare records software or by completion of standardized spreadsheets. It is subsequently linked to other healthcare databases (e.g. Hospital Episodes Statistics Database in England) to add information on hospital attendances or procedures.

This study utilized data from the 2007 to 2008 NDA cycle. The NHS Information Centre provided a pre-specified dataset including fields such as diabetes type, NHS number, Primary Care Trust, demographic data (e.g. gender, ethnicity), laboratory results (e.g. serum creatinine) and clinical measurements [e.g. body mass index (BMI)]. The National Clinical Audit Support Programme and NDA have approval to collate and analyse data from primary and secondary care under Section 251 of the NHS Act 2006 (Application No. 0353, PIAG Reference ECC 1-06/2009). Data provided for use in this study had been fully anonymized prior to analysis. No additional personal or clinical information was gathered for this study, therefore, there was no threat to patient anonymity. Therefore, separate ethical approval was not sought for this study.

Only patients with type 1 diabetes or type 2 diabetes were included in this study. The CKD Epidemiology Collaboration (CKD-EPI) equation was used to calculate estimated GFRs (eGFRs).17 Serum creatinine values provided were not initially calibrated to isotope dilution mass spectrometry.18 Therefore, creatinine values were reduced by 5% in accordance with previously published methodology.19 Where ethnicity was missing, eGFRs were calculated assuming that the patient was not of black ethnicity. Albuminuria status was defined on the basis of urinary albumin:creatinine ratio (ACR) as: normoalbuminuria ACR <3 mg/mmol, microalbuminuria 3–30 mg/mmol and macroalbuminuria >30 mg/mmol. Obesity was defined as a BMI ≥30 kg/m2.

Logistic regression was used to calculate odds ratios (ORs) for the association between the presence of obesity and other variables including year of birth, year of diagnosis, diabetes type, gender, ethnicity, smoking status, Strategic Health Authority (SHA), systolic blood pressure, glycosylated haemoglobin (HbA1c), total cholesterol and CKD stage. SHAs are organizational units of the National Health Service (NHS) in England that contain many primary and secondary care providers (Figure 1). The NHS in Wales is not organized into SHAs, therefore, patients from Wales were excluded from the adjusted analysis. A separate logistic regression model with CKD as the outcome was also included in two stages. Initially, CKD stages 1–3 were coded as ‘1’ and stage 0 (normal renal function) was coded as ‘0’. Subsequently, CKD stages 4 and 5 were coded as ‘1’ and stage 0 coded as ‘0’. We used these models to examine associations between CKD stages 1–3 and 4 and 5 and the other variables described earlier, including obesity, to confirm trends identified in the primary analysis.

Figure 1.

Map of SHAs in England (adapted from Map of SHAs February 2009, NHS, UK).

Results

The initial dataset contained 1 423 669 patients. In total, 37 485 patients were excluded on the basis of diabetes type, and a further 593 624 patients were excluded from the regression model due to missing values for clinical or laboratory measurements (including 22 494 patients from Wales). Baseline characteristics of those included in the fully adjusted obesity model are shown in Table 1. As expected, type 2 diabetes patients were older with a higher mean BMI although their mean duration of diabetes was much shorter. The proportions of patients who were obese at each CKD stage are shown in Tables 2 and 3.

View this table:
Table 1

Characteristics of patients included in adjusted obesity regression model

Type 1 diabetes (n = 58 791)Type 2 diabetes (n = 733 769)
Age (year): mean (SD)49.6 (17.3)65.9 (12.4)
Gender, n (%)
    Male33 396 (56.8)408 313 (55.6)
    Female25 395 (43.2)325 456 (44.4)
Ethnicity, n (%)
    White28 110 (47.8)331 366 (45.2)
    Black1328 (2.3)22 850 (3.1)
    Asian2417 (4.1)66 856 (9.1)
    Other1080 (1.8)16 136 (2.2)
    Missing25 856 (44)296 561 (40.4)
Smoking status, n (%)
    Current smoker10 984 (18.7)94 888 (12.9)
    Ex-smoker11 024 (18.8)204 397 (27.9)
    Non-smoker (history unknown)6276 (10.7)90 003 (12.3)
    Never smoked24 815 (42.4)298 892 (40.7)
    Missing5692 (9.7)45 589 (6.2)
Duration of diabetes (year): mean (SD)18.6 (13.1)7.6 (6.5)
BMI (kg/m2): mean (SD)27.6 (5.6)30.6 (6.2)
Systolic blood pressure (mmHg): mean (SD)130.4 (16.3)134.9 (15.3)
HbA1c (%): mean (SD)8.4 (1.7)7.3 (1.4)
Serum cholesterol (mmol/l): mean (SD)4.4 (1)4.3 (1)
Estimated GFR (ml/min/1.73 m2): mean (SD)87.6 (24.8)74.8 (20.7)
View this table:
Table 2

Prevalence and ORs of obesity in type 1 diabetes

eGFR (ml/min/1.73 m2)Albuminuria statusPrevalence % within stage (n/N)OR
Unadjusted (95% CI)Adjusted (95% CI)
≥90Normal24.9 a1a1a
60–89(10 932/43 915)(ref cat)(ref cat)
45–5937.91.84*1.21*
(1249/3294)(1.71–1.98)(1.11–1.31)
30–4447.12.69*1.73*
(610/1294)(2.41–3.01)(1.53–1.97)
15–2947.92.78*1.73*
(140/292)(2.21–3.50)(1.34–2.23)
<1529.31.250.83
(17/58)(0.71–2.20)(0.43–1.60)
≥90Microalbuminuria28.21.18*1.30*
(1175/4169)(1.10–1.27)(1.21–1.41)
60–8936.41.73*1.42*
(1437/3946)(1.61–1.85)(1.32–1.54)
45–5938.61.90*1.31*
(570/1477)(1.70–2.11)(1.16–1.47)
30–4442.72.25*1.54*
(393/921)(1.97–2.56)(1.33–1.79)
15–2949.22.92*2.04*
(146/297)(2.32–3.67)(1.59–2.62)
<1542.42.22*1.88*
(25/59)(1.32–3.72)(1.04–3.38)
≥90
Macroalbuminuria23.80.941.02
(479/2011)(0.85–1.05)(0.91–1.15)
60–8934.41.58*1.3*
(581/1691)(1.43–1.75)(1.16–1.46)
45–5940.32.04*1.51*
(216/536)(1.71–2.42)(1.25–1.83)
30–4442.82.26*1.66*
(179/418)(1.86–2.75)(1.34–2.05)
15–2943.92.36*1.75*
(101/230)(1.82–3.07)(1.31–2.35)
<1534.61.60*1.27
(37/107)(1.07–2.38)(0.80–2.02)
  • aDefined as normal renal function in this study.

  • *P < 0.05.

View this table:
Table 3

Prevalence and ORs of obesity in type 2 diabetes

eGFR (ml/min/1.73 m2)Albuminuria statusPrevalence % within stage (n/N)OR
Unadjusted (95% CI)Adjusted (95% CI)
≥90Normal49.9a1a1a
60–89
(224 780/450 226)(ref cat)(ref cat)
45–5944.80.82*1.24*
(37 909/84 546)(0.80–0.83)(1.22–1.27)
30–4445.10.82*1.43*
(13 925/30 876)(0.81–0.84)(1.39–1.47)
15–2948.70.951.75*
(2525/5182)(0.90–1.01)(1.65–1.86)
<1548.20.931.65*
(163/338)(0.76–1.16)(1.30–2.10)
≥90Microalbuminuria58.41.41*1.10*
(18 656/31 956)(1.38–1.44)(1.07–1.12)
60–89470.89*1.21*
(31 132/66 274)(0.87–0.90)(1.19–1.24)
45–5942.10.73*1.26*
(11 833/28 127)(0.71–0.75)(1.23–1.30)
30–4441.10.70*1.32*
(6772/16 468)(0.68–0.72)(1.28–1.37)
15–2944.30.80*1.56*
(2114/4769)(0.75–0.85)(1.47–1.66)
<1546.40.871.56*
(224/483)(0.73–1.04)(1.28–1.92)
≥90Macroalbuminuria58.11.39*1.00
(7629/13 134)(1.34–1.44)(0.97–1.04)
60–8947.50.91*1.07*
(11 705/24 667)(0.88–0.93)(1.04–1.10)
45–59440.79*1.23*
(4023/9147)(0.76–0.82)(1.17–1.28)
30–4445.20.83*1.42*
(2467/5453)(0.79–0.87)(1.34–1.51)
15–29460.86*1.47*
(1044/2268)(0.79–0.93)(1.34–1.61)
<1542.20.73*1.27*
(265/628)(0.62–0.86)(1.05–1.54)
  • aDefined as normal renal function in this study.

  • *P < 0.05.

Initial unadjusted regression models (including only CKD stage and albuminuria status) included 64 715 type 1 diabetes and 774 542 type 2 diabetes patients. Following adjustment, the final models included 58 791 type 1 diabetes and 733 769 type 2 diabetes patients. In unadjusted models, the ORs of obesity varied between type 1 diabetes and type 2 diabetes patients. Among type 1 diabetes patients, the unadjusted ORs of obesity were significantly higher in almost all CKD stages when compared to those with normal renal function. For example, the unadjusted ORs [95% confidence interval (CI)] of obesity in type 1 diabetes patients with macroalbuminuria varied from 0.94 (0.85–1.05) in those with an eGFR ≥90 ml/min/1.73 m2 to 1.6 (1.07–2.38) in those with an eGFR <15 ml/min/1.73 m2 (Table 2). However, among type 2 diabetes patients, the unadjusted ORs of obesity were lower in CKD stages 3–5 when compared with normal renal function, e.g. in patients with macroalbuminuria unadjusted ORs (95% CI) of obesity varied from 1.39 (1.34–1.44) in those with an eGFR ≥90 ml/min/1.73 m2 to 0.73 (0.63–0.86) in those with an eGFR <15 ml/min/1.73 m2 (Table 3).

Following adjustment, there was a significant shift in the ORs of obesity (Tables 2 and 3; Figures 2 and 3). At almost all CKD stages and degrees of albuminuria, the ORs of obesity were elevated. For example, in type 2 diabetes patients with macroalbuminuria, the adjusted ORs (95% CI) of obesity varied from 1 (0.97–1.04) in those with an eGFR ≥90 ml/min/1.73 m2 to 1.27 (1.05–1.54) in those with an eGFR <15 ml/min/1.73 m2 (Table 3). When re-analysed with CKD stages 1–3 as the outcome, the adjusted odds of CKD stage 1–3 (compared with normal renal function) were increased in obese patients compared with non-obese patients (OR 1.35, 95% CI 1.30–1.41) in type 1 diabetes patients and type 2 diabetes patients (OR 1.22, 95% CI 1.21–1.23). The corresponding adjusted ORs (95% CI) of CKD stages 4 and 5 associated with obesity were 1.93 (1.67–2.22) and 1.77 (1.72–1.86) in type 1 diabetes and type 2 diabetes, respectively.

Figure 2.

Adjusted ORs (95% CI) of obesity in type 1 diabetes by CKD stage.

Figure 3.

Adjusted ORs (95% CI) of obesity in type 2 diabetes by CKD stage.

Ethnicity was significantly associated with the odds of obesity. For example, following adjustment both type 1 diabetes and type 2 diabetes patients of Asian ethnicity were less likely to be obese (ORs 0.77 and 0.28, respectively) when compared to those of white ethnicity. However, among patients of black ethnicity divergent effects on the odds of obesity were evident between type 1 diabetes and type 2 diabetes patients. Type 1 diabetes patients of black ethnicity had a higher OR (95% CI) of 1.51 (1.33–1.72), whereas type 2 diabetes patients had a lower OR of 0.67 (0.65–0.69) (Table 4). This was reflected by differences in mean BMI in each of these groups. Mean [standard deviation (SD)] BMI was 27.7 (5.7) kg/m2 in type 1 diabetes patients and 31.1 (6.3) kg/m2 in type 2 diabetes patients of white ethnicity. This compared to 27 (5.2) kg/m2 in type 1 diabetes and 28.2 (5.2) kg/m2 in type 2 diabetes patients of Asian ethnicity. Black patients with type 1 diabetes had a mean (SD) BMI of 29.2 (5.8) kg/m2, whereas those with type 2 diabetes had a mean (SD) BMI of 30.3 (5.8) kg/m2. Mean BMI values were compared using one-way analysis of variance with post hoc analyses using the Bonferroni method. The differences between mean BMI in those of white ethnicity when compared with those of Asian or black ethnicity were all highly statistically significant (P < 0.001).

View this table:
Table 4

Adjusted ORs of obesity in type 1 and 2 diabetes

OR of obesity (95% CI)
Type 1 diabetesType 2 diabetes
Female gender1.34*1.67*
(1.29–1.39)(1.66–1.69)
Year of birth0.99*1.06*
(0.99–0.99)(1.05–1.06)
Year of diagnosis1.02*1.01*
(1.02–1.02)(1.01–1.01)
Ethnicity
 White11
(ref)(ref)
 Black1.51*0.67*
(1.33–1.72)(0.65–0.69)
 Asian0.77*0.28*
(0.69–0.85)(0.27–0.28)
 Other1.090.56*
(0.94–1.26)(0.55–0.58)
 Missing0.980.9*
(0.94–1.02)(0.89–0.91)
SHA
 London11
(ref)(ref)
 East Midlands1.34*1.25*
(1.23–1.45)(1.23–1.28)
 East of England1.13*1.15*
(1.04–1.23)(1.12–1.17)
 North East1.44*1.34*
(1.31–1.58)(1.31–1.38)
 North West1.19*1.18*
(1.1–1.28)(1.16–1.21)
 South Central1.091.17*
(1–1.2)(1.15–1.2)
 South East Coast1.111.16*
(1–1.2)(1.13–1.18)
 South West1.11*1.2*
(1.02–1.21)(1.17–1.22)
 West Midlands1.47*1.24*
(1.36–1.59)(1.21–1.27)
 Yorkshire and Humber1.2*1.33*
(1.11–1.31)(1.31–1.36)
Smoking status
 Never smoked11
(ref)(ref)
 Non-smoker (history unknown)1.12*1.18*
(1.05–1.19)(1.17–1.2)
 Ex-smoker1.24*1.27*
(1.18–1.3)(1.25–1.28)
 Current smoker0.68*0.75*
(0.64–0.72)(0.74–0.76)
 Missing0.87*0.9*
(0.81–0.93)(0.89–0.91)
Renal replacement therapy0.67*0.92
(0.46–0.97)(0.8–1.07)
Retinal photocoagulation1.37*1.02
(1.16–1.62)(0.93–1.12)
Minor amputation1.20.81*
(0.81–1.78)(0.68–0.96)
Major amputation0.740.82
(0.36–1.52)(0.62–1.07)
Systolic blood pressure1.02*1.01*
(1.02–1.02)(1.01–1.01)
HbA1c11.09*
(0.99–1.02)(1.09–1.09)
Serum cholesterol0.96*0.93*
(0.95–0.98)(0.92–0.93)

Regional variations in the odds of obesity were also evident. Among type 1 diabetes patients only the South Central and South East Coast SHAs did not have higher ORs of obesity, whereas among type 2 diabetes patients being resident in any SHA outside London was associated with higher ORs of obesity. ORs for other variables included in the analysis are shown in Table 4.

Discussion

This study has demonstrated a strong association between kidney disease (defined as an eGFR <60 ml/min/1.73 m2, albuminuria or both) and obesity (defined as a BMI ≥30 kg/m2) in both type 1 diabetes and type 2 diabetes in a large English diabetes population. We also identified various other risk factors for the development of obesity including ethnicity, SHA and smoking status (Table 4).

Obesity is a frequent finding in type 2 diabetes patients, however, Conway et al. recently documented a dramatic increase in the incidence of obesity among type 1 diabetes patients in an American cohort.20 In the Diabetes Control and Complications Trial (type 1 diabetes patients recruited from 1983 to 1989), mean BMI rose over 22 years of follow-up from 23.5 kg/m2 at inception to 29.4 kg/m2 in the intensive therapy arm and 28.2 kg/m2 in the control arm.21 Although this could be an age-related change it may also reflect trends towards increased obesity rates in the general population. As shown in Tables 2 and 3, there were marked differences between the unadjusted and adjusted ORs of obesity associated with DKD. Although this shift in the adjusted ORs resulted from the adjustment for multiple variables, the effect of gender and age (year of birth in this analysis) accounted for most of the difference. Females were much more likely to be obese than males and this reflects trends documented by the WHO in many populations worldwide.22 Divergent effects of aging on the ORs of obesity were also notable with younger type 1 diabetes patients having lower ORs as opposed to higher ORs in younger type 2 diabetes patients. Increasing rates of obesity have been documented in younger age groups in population-based studies.23 The higher odds of obesity in younger type 2 diabetes patients in this study could relate to increased rates of obesity at younger ages in the general population. If patients develop obesity at a younger age then they conceivably would also develop the associated metabolic changes, such as insulin resistance, which predispose to type 2 diabetes at an earlier age. Our study has also confirmed that, as kidney disease declines, patients with type 1 diabetes as well as type 2 diabetes are increasingly likely to be obese when compared to those patients with normal renal function. Overall, 28.4% of type 1 diabetes patients and 48.8% of type 2 diabetes patients were obese in this study. Obesity is associated with increased risks of many other conditions such as cardiovascular disease or chronic respiratory disorders.24,25 The additional mortality risk associated with these complications, in combination with DKD-associated mortality risk, may help to explain why relatively few DKD patients survive to reach end-stage renal disease.

We also explored the relationship between the odds of having DKD in the presence of obesity and found that obesity was associated with the presence of all stages of DKD. Previous studies in both type 1 diabetes and type 2 diabetes have documented conflicting results with respect to the effect of obesity on DKD.26–28 Our study is cross-sectional in nature and, therefore, it is difficult to imply causation as both renal function and BMI are measured simultaneously. However, it is striking that, even at low levels of eGFR, a strong association persists between DKD and obesity in both diabetes types. Further research is clearly needed to further examine these relationships and identify any potential causal pathways.

Our study also demonstrated that the association between obesity and DKD is modified by other factors, such as ethnicity and UK region. Surprisingly, our results suggest that Asian patients with both type 1 diabetes and type 2 diabetes, regardless of level of kidney function, are less likely to be obese than those of white ethnicity. However, this reflects both the use of BMI as a measure of obesity and differing body fat and muscle contents in ethnic groups.29–34 It is likely that current ‘cut-off’ points for defining obesity are not representative of cardiovascular risk in certain ethnic groups.31,32,34 Some studies have suggested that abnormalities associated with the metabolic syndrome, such as hyperglycaemia or dyslipidaemia, occur at much lower BMI values in Asian patients (as low as 21 kg/m2).34 Arguably, BMI therefore should be interpreted in an ethnicity-specific manner.

Socioeconomic status is also associated with the risk of developing cardiovascular disease and type 2 diabetes. Although we could not adjust at an individual-patient level for socioeconomic status there were clear regional variations in risk of obesity when analysed by SHA. Holding all other variables in the model constant (including CKD stage, gender and ethnicity), the ORs of obesity were higher in regions outside London, the South East or South Coast. These differences in odds of obesity may partly reflect differing socioeconomic status between regions or, alternatively, could reflect variations in healthcare provision or public health strategies. Unfortunately, we were unable to clarify this using the data available, and further research is clearly needed to investigate the causes of these regional variations.

The major strengths of this study are its large size and the diverse geographical and socioeconomic backgrounds of the patients included. The NDA is reported to be the largest clinical audit in the world and collates detailed information on many aspects of diabetes care. This allowed us to construct a detailed regression model with sufficient power to detect even small increments in ORs for multiple co-variates. However, there are some potential weaknesses. First, the data were cross-sectional which limited its interpretation and extrapolation beyond this audit cycle. Arguably, many clinical and laboratory measurements should not be defined by single measurements, e.g. blood pressure.35 CKD, in itself, should not be defined on the basis of single measurements of eGFR or albuminuria but there is precedence for this in previous studies.36 Second, participation in the NDA is voluntary and, in 2007–08, participation in some regions was poor. This could introduce selection bias into the sample. However, if this were the case, it would seem likely that this would positively bias results (i.e. towards lower BMIs) as poorer performing primary and secondary care providers would be less likely to contribute. Finally, the use of BMI as a measure of nutritional status is difficult in certain ethnic groups, as described earlier, and also in CKD. More advanced CKD is associated with impaired sodium and free water excretion leading to volume overload. This excess fluid and thus weight could artificially raise a patient’s BMI. This has led to attempts to find better measurements of body fat and composition but none are currently in routine clinical practice.37

In conclusion, this study has identified a strong association between obesity and kidney disease in type 1 diabetes. It has also confirmed a similar association in type 2 diabetes. If causal in nature, then the rising prevalence of obesity among type 1 diabetes patients could predict a future increase in kidney disease among these patients. The ethnic and regional disparities we have described may also help to inform public health policy and direct future targeted weight reduction initiatives.

Funding

This work was supported by the Northern Ireland Kidney Research Fund and NHS Kidney Care.

Conflict of interest: None declared.

Acknowledgements

The authors would like to acknowledge Julie Henderson, AlaUddin and colleagues in the NHS Information Centre as well as all the contributing practices and hospital units. Thanks to Dr Rowan Hilson and James Metcalfe for supportive discussions on the NDA. The funding body had no role in the collection, analysis or interpretation of data included in this study.

References

View Abstract