The authors have declared that no competing interests exist.
Conceived and designed the experiments: APK GDB SC. Analyzed the data: ES APK GDB SC. Wrote the paper: APK. Critical revision of the manuscript: MH ES GDB SC.
Both anaemia and cardiovascular disease (CVD) are common in people with diabetes. While individually both characteristics are known to raise mortality risk, their combined influence has yet to be quantified. In this pooling project, we examined the combined impact of baseline haemoglobin levels and existing CVD on all-cause and CVD mortality in people with diabetes. We draw comparison of these effects with those apparent in diabetes-free individuals.
A combined analyses of 7 UK population-based cohorts resulted in 26,480 study members. There were 946 participants with physician-diagnosed diabetes, 2227 with anaemia [haemoglobin<13 g/dl (men) or <12 (women)], 2592 with existing CVD (stroke, ischaemic heart disease), and 21,396 with none of the conditions. Across diabetes and anaemia subgroups, and using diabetes-free, non-anaemic participants as the referent group, the adjusted hazard ratios (HR) were 1.46 (95% CI: 1.30–1.63) for anaemia, 1.67 (1.45–1.92) for diabetes, and 2.10 (1.55–2.85) for diabetes and anaemia combined. Across combined diabetes, anaemia and CVD subgroups, and compared with non-anaemic, diabetes-free and CVD-free participants, HR (95% CI) for all-cause mortality were 1.49 (1.32–1.69) anaemia, 1.60 (1.46–1.76) for existing CVD, and 1.66 (1.39–1.97) for diabetes alone. Equivalents were 2.13 (1.48–3.07) for anaemia and diabetes, 2.68 (2.14–3.36) for diabetes and existing CVD, and 3.25 (1.88–5.62) for the three combined. Patterns were similar for CVD mortality.
Individually, anaemia and CVD confer similar mortality risks in people with diabetes, and are excessively fatal in combination. Screening for anaemia would identify vulnerable diabetic patients whose outcomes can potentially be improved.
Anaemia is frequent in people with diabetes where it is generally undetected and therefore untreated
CVD is common in people with diabetes
Participants were 26,480 individuals with data available on diabetes status (history of physician-diagnosed) and total haemoglobin level at baseline
The full study protocol has been described in detail elsewhere
Systolic and diastolic blood pressure was measured with an Omron HEM-907 blood pressure monitor three times in the sitting position after 5-min rest between each reading. The average of the second and third BP recordings was used for the present analyses. Height and weight were measured directly by the interviewers using Chasmors stadiometers (Chasmors Ltd, London, UK) and Tanita electronic digital scales (Tanita, Corporation, Tokyo, Japan), respectively. BMI was calculated using the usual formulae (weight [kg]/height [m2]). Waist and hip circumferences were measured using a tape with an insertion buckle at one end. Waist circumference was measured at the midpoint between the lower rib and the upper margin of the iliac crest. Hip circumference was denoted by the widest circumference around the buttocks, below the iliac crest. Both measurements were taken twice, using the same tape, and were recorded to the nearest even millimetre. Those whose two waist or hip measurements differed by more than 3 cm had a third measurement taken. The mean of the two valid measurements was used in our analysis. Cholesterol was measured using cholesterol oxidase assays on an Olympus 640 analyzer. Anaemia was defined as haemoglobin concentrations <13 g/dl (men) and <12 g/dl (women), following the World Health Organisation criteria
Consenting participants were linked to UK National Health Service records from which a death certificate was located. Classification of the underlying cause of death was based on information on the death certificate together with any additional observations made by the certifying physician. Diagnoses for primary cause of death used the ninth (ICD-9) and tenth (ICD-10) revisions of the International Classification of Diseases. Cardiovascular disease codes were 390–459 for ICD-9 and I01–I99 for ICD-10.
The starting study sample comprised 57,073 participants, among whom 28,809 (46.2%) had provided blood sample for total haemoglobin assays. Sixteen were excluded for missing data on diabetes status. Other 2313 participants who did not consent for mortality follow-up were also excluded. Therefore primary analyses were based on 26,480 individuals (12,135 men) with data available on age, sex, diabetes status and haemoglobin level at baseline (
Participants were classified according anemia and prior CVD status, then further grouped by baseline diabetes status. Baseline comparisons used logistic regressions and generalized linear regression models. A Poisson model was used to determine the absolute risk of CVD and all-cause mortality during follow-up by status for anaemia, and for prior CVD. Kaplan-Meier estimator was used to compute the probability of death during follow-up and estimates compared across baseline stratification variables with the use of the Log-Rank test. Cox regression models were used to investigate the associations of anaemia and prior CVD with mortality after adjustment for cohort, age, sex, smoking, systolic blood pressure, body mass index and total cholesterol.
To investigate the association between total haemoglobin and mortality risks, Cox models were used to compute the hazard ratio and accompanying 95% confidence interval (95% CI) for a one standard deviation (SD) decrease in total haemoglobin in relation to all-cause and CVD mortality. Similar Cox models were used to compare mortality risk across fifths of haemoglobin, with 95% CI derived with the used of floating absolute risk methods
Among eligible participants, 13,228 (424 with diabetes, 3.3%) had data available on CRP levels. Multivariable Cox regression analyses were conducted in this subgroup with further adjustment for CRP levels to investigate the potential effects of chronic inflammation on the observed results.
The study sample included 26,480 participants among whom 946 (3.6%) had diabetes. The prevalence of anaemia was higher (14.3%; n = 135) in participants with diabetes than those without (8.2%; n = 2092; p-value for difference <0.001). Compared to their non-anaemic counterparts, participants with anaemia were more likely to be female, to smoke, and have with a history of CVD. They were also older but had lower systolic blood pressure, body mass index, waist circumference and total cholesterol (
Variables | No diabetes | Diabetes | All participants | |||||
no anaemia | Anaemia | p-value | no anaemia | anaemia | p-value | Anaemia (yes vs. no) | Diabetes (yes vs. no) | |
N | 23442 | 2092 | 811 | 135 | ||||
Women (%) | 53.1% | 70% | <0.001 | 44% | 49.6% | 0.22 | <0.001 | <0.001 |
Mean age, years (SD) | 54.4 (13.2) | 59.0 (17.1) | <0.001 | 61.9 (11.9) | 68.8 (14.2) | <0.001 | <0.001 | <0.001 |
Current smoking (%) | 27.4% | 15.2% | <0.001 | 24.3% | 11.9% | 0.001 | <0.001 | 0.008 |
Mean systolic bloodpressure*, mmHg (SD) | 137 (20) | 134 (22) | <0.001 | 146 (21) | 145 (25) | 0.57 | <0.001 | <0.001 |
Mean resting heart rate,bpm† (SD) | 71 (11) | 71 (11) | 0.13 | 74 (12) | 74 (10) | 0.65 | 0.06 | <0.001 |
Mean body mass index,kg/m2 (SD) | 27.0 (4.5) | 25.8 (4.8) | <0.001 | 29.1 (5.1) | 26.8 (4.1) | <0.001 | <0.001 | <0.001 |
Mean waist circumference,cm (SD) | 90.0 (12.9) | 85.7 (12.4) | <0.001 | 98.6 (13.1) | 94.1 (11) | <0.001 | <0.001 | <0.001 |
Mean waist/hip ratio (SD) | 0.86 (0.09) | 0.84 (0.08) | <0.001 | 0.92 (0.08) | 0.91 (0.07) | 0.10 | <0.001 | <0.001 |
Mean total cholesterol,mmol/l (SD) | 6.0 (1.2) | 5.5 (1.1) | <0.001 | 5.8 (1.1) | 5.3 (1.2) | <0.001 | <0.001 | <0.001 |
Median CRP, mg/l (25th–75th percentiles) | 1.8 (0.8–3.9) | 1.6 (0.6–4.6) | <0.001 | 3.1 (1.5–6.4) | 3.5 (1.0–16.6) | 0.007 | <0.001 | <0.001 |
Existing cardiovasculardisease (%) | 8.7% | 14.3% | <0.001 | 26.6% | 31.1% | 0.28 | <0.001 | <0.000 |
SD, standard deviation.
During follow-up, 4643 deaths from all cause were recorded, of which 1347 (29%) were from cardiovascular disease. The number of fatal outcomes in participants with diabetes was 378 for all-cause mortality and 133 for cardiovascular mortality.
The absolute risk (95% CI) per 1000 person-years for participants with and without diabetes, and by status for anaemia at baseline is shown in
Baseline classification | CVD mortality | All-cause mortality | |||||
Diabetes | Anaemia | Event rate(/1000 pys) | HR (95% CI) |
HR (95% CI) |
Event rate(/1000 pys) | HR (95% CI) |
HR (95% CI) |
No | No | 4.2 (4.0–4.5) | 1 | 1 | 14.8 (14.4–15.3) | 1 | 1 |
No | Yes | 9.3 (8.7–11.7) | 1.48 (1.20–1.81) | 1.53 (1.24–1.88) | 33.5 (31.0–36-2) | 1.44 (1.29–1.60) | 1.46 (1.30–1.63) |
Yes | No | 15.2 (12. 6–18.3) | 2.32 (1.84–2.94) | 2.00 (1.57–2.53) | 41.8 (37.3–46.9) | 1.76 (1.53–2.02) | 1.67 (1.45–1.92) |
Yes | Yes | 29.0 (19.6–42.9) | 2.08 (1.17–3.70) | 1.96 (1.10–3.50) | 90.5 (72.5–113.0) | 2.22 (1.64–3.01) | 2.10 (1.55–2.85) |
Cox models are adjusted for cohort, age, sex,
Cox models are further adjusted for smoking systolic blood pressure, total cholesterol, BMI and prior CVD.
CI, confidence interval; HR, hazard ratio; prs, person-years.
The probability of survival during follow-up is depicted in
The upper figure panel is for cardiovascular disease and the lower for all-cause mortality.
Using nondiabetics without anaemia as a reference group, anaemia without diabetes was associated with 53% (24–88%) higher risk of CVD mortality after adjustment for age, sex, prior CVD, systolic blood pressure, current smoking, body mass index and total cholesterol. Equivalents were 100% (57–153%) for diabetes without anaemia and 96% (10–250%) for diabetes & anaemia (
The absolute risk of all-cause and CVD mortality (per 1000 person-years) by crossed status for anaemia and existing CVD is summarised in
+ denotes the presence of the characteristic, and – denotes its absence.
In age, sex and cohort adjusted analysis, there was a weak positive association between haemoglobin and all-cause mortality [hazard ratio per SD lower haemoglobin: 1.07 (95% CI: 1.03–1.10), with no significant heterogeneity by diabetes status (p = 0.08 for interaction). No continuous association was found for CVD mortality. The shape of the associations for different coding of total haemoglobin is depicted in
The solid curve depicts the shape of the shape of the association across the continuum of total haemoglobin, and the shaded area if for the 95% confidence interval around the curve.
A total of 64 participants (5 with diabetes) had haemoglobin levels within the range for defining polycythaemia (i.e. haemoglobin≥18.5 g/dl in men or haemoglobin≥16.5 g/dl). Among them 30 deaths (cumulative incidence 46.9%) were recorded, 5 (cumulative incidence 7.8%) being of cardiovascular origin. Using participants with normal range haemoglobin levels as a reference group, the age and sex adjusted hazard ratios (95% confidence intervals) associated with all-cause mortality were 1.46 (1.34–1.59) for anaemia and 2.70 (1.89–3.88) for polycythaemia. The equivalents for CVD mortality were 1.28 (1.10–1.50) and 1.56 (0.65–3.77).
Among participants with data available on CRP levels (13,228 participants), 1761 (136 in people with diabetes) deaths were recorded during follow-up, of which 590 (61 in people with diabetes) were from cardiovascular disease. Across diabetes and anaemia strata, using nondiabetics without anaemia as a reference group, the age and sex adjusted HR (95% CI) associated with CVD mortality were 1.55 (1.17–2.04) for anaemia without diabetes, 2.68 (2.01–3.58) for diabetes without anaemia and 2.23 (1.10–4.53) for both anaemia and diabetes. The equivalents after further adjustment for prior CVD, systolic blood pressure, current smoking, body mass index, total cholesterol and log(CRP) were 1.43 (1.08–1.90), 2.13 (1.58–2.86) and 1.81 (0.90–3.70). For all-cause mortality, estimates were 1.39 (1.18–1.64), 1.83 (1.51–2.21) and 2.20 (1.44–3.36) in sex age adjusted models, and 1.24 (1.05–1.47), 1.65 (1.36–2.01) and 1.76 (1.15–2.71) after adjusted for all covariates including CRP. The small number of participants in some subgroups hampered our ability to reliably perform similar analysis across diabetes, anaemia and prior CVD strata. Estimates however were mostly similar to those from the main analysis (
In this pooling of contemporary, community-based cohort studies, we found that anaemia was associated with increased risks of all-cause and cardiovascular disease mortality. The magnitude of these risks in people with diabetes and no history of CVD were similar to those conferred by a history of CVD. In diabetic participants with existing CVD, anaemia conveyed very high mortality risk. There was no continuous linear association between haemoglobin levels and mortality risk. Both lower and higher haemoglobin levels were associated with higher mortality risks.
Previous reports have separately examined the effects of anaemia or existing CVD on the risk of major outcomes. Investigations on the effects of anaemia have mostly focused on people with chronic kidney disease and have largely confirmed the related high risk of mortality
We found that in both people with and without diabetes, anaemia was associated with increased risk of mortality in participants with prior CVD. In accordance with our results, previous population-based studies found that the association between haemoglobin levels and cardiac events, if any, was not linear
Anaemia in diabetes has been attributed to erythropoietin (EPO) deficiency subsequent to renal complications
Trials of anaemia correction to reduce cardiovascular risk conducted so far, exclusively in people with CKD, support a harmful effect of anaemia correction particularly when erythropoiesis-stimulating agents are used to raise haemoglobin levels into ‘normal range’
Our study has some limitations. We lacked data on kidney function and could not account for possible effects of nephropathy
In conclusion, anaemia is a determinant of all-cause and cardiovascular mortality. Diabetic individuals with anaemia but no prior CVD have risks of mortality similar to those among CVD survivors with diabetes, but no anaemia. The previous focus on EPO for anaemia correction still leaves unaddressed other determinants of chronic anaemia in diabetes. Addressing these determinants and non-optimal cardiovascular risk profile may improve the outcomes of patients. To be effective however, interventions should be implemented early in routine diabetes care, and not at the kidney impairment stage
(TIF)
(TIF)
Profile of participants included and those excluded.
(DOC)
Baseline characteristics by status for anaemia and existing cardiovascular disease (CVD) in participants with and without diabetes.
(DOC)
Baseline characteristics across fifths of total haemoglobin according to diabetes status.
(DOC)
Unadjusted Incidence of Cardiovascular and all-cause mortality per 1000 person-years of follow-up and adjusted hazard ratio by status for diabetes, anaemia and existing cardiovascular disease.
(DOC)
Fit statistics for various coding of total haemoglobin in relation with all-cause and cardiovascular mortality risk.
(DOC)
We would like to thank the Health Survey for England and Scottish Health Survey respondents for offering their valuable time and the Information Services Division Scotland team for their outstanding Scottish Health Survey data provision and data updating services.