clinical investigation
http://www.kidney-international.org & 2013 International Society of Nephrology
see commentary on page 10
Subclinical cardiovascular disease is associated with a high glomerular filtration rate in the nondiabetic general population Bjørn O. Eriksen1,2,3, Maja-Lisa Løchen4, Kjell A. Arntzen1,5, Geir Bertelsen1,6, Britt-Ann W. Eilertsen3, Therese von Hanno1,7, Marit Herder4,8, Trond G. Jenssen1,9, Ulla D. Mathisen2, Toralf Melsom2, Inger Njølstad4, Marit D. Solbu2, Ingrid Toft1,2 and Ellisiv B. Mathiesen1,5 1
Department of Clinical Medicine, Faculty of Health Sciences, University of Tromsø, Tromsø, Norway; 2Section of Nephrology, University Hospital of North Norway, Tromsø, Norway; 3Department of Clinical Research, University Hospital of North Norway, Tromsø, Norway; 4 Department of Community Medicine, Faculty of Health Sciences, University of Tromsø, Tromsø, Norway; 5Department of Neurology and Neurophysiology, University Hospital of North Norway, Tromsø, Norway; 6Department of Ophthalmology, University Hospital of North Norway, Tromsø, Norway; 7Department of Ophthalmology, Nordland Hospital, Bodø, Norway; 8Department of Radiology, University Hospital of North Norway, Tromsø, Norway and 9Department of Nephrology, Oslo University Hospital, Oslo, Norway
A reduced glomerular filtration rate (GFR) in chronic kidney disease is a risk factor for cardiovascular disease. However, evidence indicates that a high GFR may also be a cardiovascular risk factor. This issue remains unresolved due to a lack of longitudinal studies of manifest cardiovascular disease with precise GFR measurements. Here, we performed a cross-sectional study of the relationship between high GFR measured as iohexol clearance and subclinical cardiovascular disease in the Renal Iohexol Clearance Survey in Tromsø 6 (RENIS-T6), a representative sample of the middle-aged general population. A total of 1521 persons without cardiovascular disease, chronic kidney disease, diabetes, or micro- or macroalbuminuria were examined with carotid ultrasonography and electrocardiography. The GFR in the highest quartile was associated with an increased odds ratio of having total carotid plaque area greater than the median of non-zero values (odds ratio 1.56, 95% confidence interval 1.02–2.39) or electrocardiographic signs of left ventricular hypertrophy (odds ratio 1.62, 95% confidence interval 1.10–2.38) compared to the lowest quartile. The analyses were adjusted for cardiovascular risk factors, urinary albumin excretion, and fasting serum glucose. Thus, high GFR is associated with carotid atherosclerosis and left ventricular hypertrophy and should be investigated as a possible risk factor for manifest cardiovascular disease in longitudinal studies. Kidney International (2014) 86, 146–153; doi:10.1038/ki.2013.470; published online 4 December 2013 KEYWORDS: arteriosclerosis; cardiovascular disease; glomerular filtration rate; hyperfiltration; left ventricular hypertrophy; risk factors Correspondence: Bjørn O. Eriksen, Section of Nephrology, University Hospital of North Norway, 9038 Tromsø, Norway. E-mail: [email protected] Received 7 June 2013; revised 20 August 2013; accepted 19 September 2013; published online 4 December 2013 146
Low glomerular filtration rate (GFR) is a risk factor for cardiovascular disease (CVD).1 Although high GFR has been regarded as normal, evidence suggests that it may also confer increased risk. In a pooled analysis of over two million subjects, Nitsch et al.2 found an increased cardiovascular mortality for nonproteinuric men with high GFR, whereas a lower increase for women did not reach statistical significance. These analyses were adjusted for the presence of diabetes. Another study found an increased risk of coronary heart disease with high GFR in individuals without prevalent CVD, diabetes, or chronic kidney disease.3 As these studies used serum creatinine to estimate GFR, it has been assumed that these and similar findings can be explained by an association between chronic disease and low muscle mass that would result in overestimated GFR. However, this assumption has not been investigated. Because precise GFR measurements are costly and cumbersome, there are, to our knowledge, no longitudinal population-based studies of measured GFR and CVD. GFR estimates based on creatinine or cystatin C are known to be inaccurate in the upper range.4 In addition, creatinine and cystatin C are influenced by non-GFR factors, including cardiovascular risk factors, that may cause confounding.5–7 Accordingly, whether high GFR is a CVD risk factor remains unresolved. However, the hypothesis implies that subclinical CVD may be cross-sectionally associated with high GFR. We investigated this issue in the Renal Iohexol Clearance Survey in Tromsø 6 (RENIS-T6) that has the largest population-based cohort with GFR measurements. The cohort consists of a representative sample of the middle-aged general population without manifest CVD, diabetes, or chronic kidney disease, and who underwent carotid ultrasonography and electrocardiography (ECG). Carotid atherosclerosis and left ventricular hypertrophy (LVH) are signs of subclinical CVD that predict manifest CVD.8–10 Our aim was to investigate whether Kidney International (2014) 86, 146–153
clinical investigation
BO Eriksen et al.: Subclinical cardiovascular disease and high GFR
there is a cross-sectional association between subclinical CVD and high GFR in apparently healthy individuals. RESULTS
Of the 1627 subjects included in the RENIS-T6, individuals with GFR o60 ml/min per 1.73 m2 (n ¼ 34), micro- or macroalbuminuria (n ¼ 44), or diabetes (n ¼ 33) were excluded. Three individuals had both diabetes and microalbuminuria; two had both diabetes and GFR o60 ml/min per 1.73 m2. Accordingly, 106 subjects were excluded, leaving 1521 for the present analyses (Figure 1). There were differences across the GFR quartiles with regard to age; gender; height; body weight; current smoking; ambulatory systolic blood pressure (BP), diastolic BP, mean arterial pressure, and pulse pressure; antihypertensive drug use; high-density lipoprotein cholesterol and fasting glucose (Po0.05) (Table 1). These findings have been discussed in previous publications.5,11,12 In unadjusted analyses, higher GFR quartile was associated with greater carotid mean intima media thickness (IMT), a higher number of carotid plaques, greater total plaque area (TPA), greater Sokolow and Multi-Ethnic Study of Atherosclerosis (MESA) voltage, and a higher percentage of LVH detected by the MESA voltage criterion (Po0.05 for N = 5464 Invited to the Tromsø 6 Study and aged 50 to 62 years
N = 3564 Met and completed the main Tromsø 6 Study
N = 2825 Invited to RENIS-T6
N = 2107 Responders
N = 739 Reported a previous myocardial infarction, angina pectoris, stroke, diabetes mellitus, or any renal disease except urinary tract infection.
N = 125 Excluded because of allergy to contrast media, iodine, or latex, or for other reasons. Includes 48 who withdrew.
N = 1982 Eligible for inclusion
N = 1632 Investigated in RENIS-T6 according to a predetermined target
N = 1627 The RENIS-T6 cohort
N = 1521 Present study population
N=5 Technical failure in the iohexol-clearance measurements. N = 106 Diabetes (N = 33) according to their fasting plasma glucose or HbA1c. CKD (N = 34) defined as measured GFR< 60 ml/min/m or micro- or macroalbuminuria (N = 44) (3 persons had both diabetes and microalbuminuria, 2 both diabetes and CKD)
Figure 1 | Study population derived from RENIS-T6, an ancillary investigation of the sixth Tromsø Study. CKD, chronic kidney disease; GFR, glomerular filtration rate; RENIS-T6, Renal Iohexol Clearance Survey in Tromsø 6. Kidney International (2014) 86, 146–153
linear trend; Table 2). There was an opposite trend for the Cornell product (Table 2). In the calculation of the Cornell product, a factor of 0.8 is added for women, who have a lower GFR than men, which explains this finding. Quadratic trends were not statistically significant for any of the variables in Table 2. More information about the relationships between covariates and the outcomes can be found in Supplementary Tables S4 to S6 online. In the adjusted models (Table 3), the odds ratios (ORs) for the dichotomous IMT and TPA variables increased with increasing quartile of GFR (Po0.05). The ORs for the highest quartile compared with the lowest quartile were 1.49 for IMT and 1.56 for TPA in the fully adjusted models (model 4; Po0.05). Both IMT and TPA were also associated with GFR as a continuous variable (Po0.05). In the adjusted models, LVH was judged to be present when detected by either the Sokolow or MESA voltage criterion, or the Cornell product criterion. In all the adjusted models (Table 3), the OR for LVH was elevated for the highest compared with the lowest GFR quartile (Po0.05). For the fully adjusted model, the OR was 1.62. When LVH by each criterion was analyzed separately, OR was increased in the highest GFR quartile in all the adjusted models for the Cornell product criterion (Po0.05), and there was a linear trend for increasing OR with increasing GFR for the MESA voltage in the fully adjusted model (Po0.05)(Supplementary Table S2 online). LVH as detected by the Sokolow voltage criterion was not associated with GFR. There were positive associations between both the Sokolow voltage and MESA voltage analyzed as continuous variables and GFR (Po0.05), but there was no association with the Cornell product (Supplementary Table S3 online). None of the interactions between sex and GFR in any of the fully adjusted models in Table 3 were statistically significant. When analyzing fractional polynomial models with the same dependent variables as in Table 3 and the same independent variables as in the fully adjusted models (model 4), no nonlinear effects of GFR were found (PX0.05). To explore possible confounding by body size, all the multiple regression analyses were repeated with GFR standardized to a total body water of 40 l (GFR40) as the dependent variable instead of GFR standardized to body surface area. The results of these analyses were similar, as were also the results of analyses using absolute GFR in ml/min (not shown). DISCUSSION
To our knowledge, this investigation is the first to find an independent association between high normal GFR and subclinical CVD manifested as carotid atherosclerosis and LVH in the healthy general population. The cross-sectional relationship between GFR and carotid atherosclerosis has been examined previously. Several studies have found no statistically significant association between GFR and IMT,13,14 whereas two studies found an association 147
clinical investigation
BO Eriksen et al.: Subclinical cardiovascular disease and high GFR
Table 1 | Study population characteristics according to quartile of glomerular filtration rate (the RENIS-T6 study) Quartile of glomerular filtration rate, range (ml/min per 1.73 m2)
Glomerular filtration rate, ml/min per 1.73 m2 Female gender, N (%) Age, years Height, cm Body weight, kg Body mass index, kg/m2 First-degree relative with myocardial infarction before 60 years, N (%) Physical exercisea, N (%) Current smoking, N (%) Mean no. of cigarettes smoked per day, if currently smoking LDL cholesterol, mmol/l HDL cholesterol, mmol/l Triglyceridesb, mmol/l High-sensitivity C–reactive protein, mg/l Fasting serum glucose, mmol/l Urine albumin–creatinine ratio, mg/mmol
Quartile 1 (n ¼ 380) (60.9–83.4)
Quartile 2 (n ¼ 380) (83.4–91.8)
Quartile 3 (n ¼ 381) (91.9–101.2)
Quartile 4 (n ¼ 380) (101.2–138.6)
P for linear trend
77.6 (72.8–80.8) 256 (67.4) 60.5 (56.6–62) 169.2 (163.1–175.5) 77.6 (68.4–88.6) 26.7 (24.4–29.7) 88 (23.2)
87.6 (85.4–89.7) 211 (55.5) 59.2 (54.7–61.5) 169.7 (164.0–177.0) 77.5 (68.1–87.1) 26.5 (24.2–29.3) 62 (16.3)
96.1 (94.0–98.7) 184 (48.3) 57.8 (54.2–61.) 169.8 (164.0–177.2) 77.9 (69.4–88.9) 26.9 (24.5–29.5) 80 (21.0)
107.6 (103.9–113.7) 120 (31.6) 57.1 (53.6–60.8) 173.0 (166.4–178.2) 81.4 (71.2–90.7) 27.0 (24.7–29.9) 72 (18.9)
o0.0001 o0.0001 o0.0001 0.02 0.95 o0.0001
195 (51.3) 73 (19.3) 10.4
228 (60.1) 63 (16.6) 11.5
228 (59.9) 73 (19.2) 12.1
197 (51.8) 111 (29.2) 11.8
0.94d o0.0001d o0.0001d
3.60 1.50 1.00 1.21 5.20 0.29
(3.00–4.20) (1.30–1.85) (0.80–1.50) (0.66–2.21) (4.90–5.50) (0.17–0.54)
3.60 1.50 1.00 1.15 5.30 0.28
(3.10–4.20) (1.30–1.80) (0.70–1.30) (0.61–2.16) (5.00–5.50) (0.16–0.52)
3.70 1.50 1.10 1.17 5.30 0.26
(3.20–4.30) (1.20–1.80) (0.70–1.50) (0.65–2.16) (5.00–5.60) (0.15–0.46)
3.60 1.40 1.10 1.18 5.30 0.29
(3.10–4.20) (1.20–1.70) (0.70–1.50) (0.64–2.08) (5.10–5.80) (0.15–0.54)
0.45 o0.0001 0.13d 0.76 o0.0001 0.63d
Daytime ambulatory blood pressure Systolic, mm Hg Diastolic, mm Hg Mean arterial, mm Hg Pulse pressure, mm Hg
127.8 (119.2–137.3) 80.5 (74.3–86.3) 97.2 (90.5–103.9) 47.0 (42.3–52.7)
128.3 (119.3–136.1) 81.5 (75.8–86.4) 98.1 (91.1–103.6) 46.3 (41.6–51.2)
129.1 (120.4–137.6) 81.9 (76.2–86.9) 98.3 (91.6–104.) 46.5 (42.2–51.9)
132.1 (123.5–140.1) 83.3 (77.7–88.3) 99.6 (93.8–105.6) 48.2 (43.1–53.0)
o0.0001 o0.0001 o0.0001 0.04
Hypertensionc, N (%) Antihypertensive medication, N (%)
171 (45.0) 81 (21.3)
160 (42.1) 62 (16.3)
159 (41.7) 64 (16.8)
145 (38.2) 58 (15.3)
0.06 0.01
Abbreviations: HDL, high-density lipoprotein; LDL, low-density lipoprotein; RENIS-T6, Renal Iohexol Clearance Survey in Tromsø 6. Data are shown as median (interquartile range) or percent. a Defined as leisure-time physical exercise leading to perspiration or breathlessness (yes/no). b Log-transformed when differences between quartiles were tested. c Defined as conventional systolic blood pressure (BP) 4140, conventional diastolic BP 490, or the use of antihypertensive medication. d Po0.05 for quadratic trend.
between low GFR and increased IMT.15,16 The conflicting results may reflect shortcomings of the GFR assessments, as all these studies used GFR estimates rather than measured GFR. Two Asian studies of GFR and plaques also showed differing results.17,18 These studies investigated only the presence or absence of plaques, not plaque area. To our knowledge, there are no population-based studies of the relationship between GFR and LVH based on ECG findings. However, MESA investigated GFR and LV mass measured with magnetic resonance imaging at baseline.19 For a cystatin C–based estimated GFR o75 ml/min per 1.73 m2, both the ORs for LVH and LV mass were increased, but a linear regression found no association for estimated GFR 475 ml/min per 1.73 m2. Because this analysis was unadjusted, it does not exclude an association between high GFR and LVH. In addition, cystatin C–based estimated GFR is known to be influenced by several non-GFR factors that may have affected the result.5–7 LVH detected by ECG is known to be confounded by body mass index. For this reason, we performed analyses in which GFR was standardized to total body water, which avoids confounding from body size.20 In addition, we repeated the analyses with absolute GFR in ml/min, which have been used by some authors to avoid underestimation of renal function 148
in obese individuals.21 The results of these analyses differed little from the analyses with GFR standardized to body surface area that indicates that body size is not a confounder of the association between LVH and GFR. The association between carotid atherosclerosis, LVH, and high normal GFR may seem paradoxical. However, associations between increased GFR and prediabetes, high blood pressure, increased pulse rate, smoking, and other cardiovascular risk factors have been found in RENIS-T6 and other studies of nondiabetic subjects.5,11,12,22–26 The mechanism for these associations has not been established, but increased activity of the sympathetic nervous system may be one contributing factor.27 In addition, one hypothesis regarding the pathogenesis of chronic kidney disease postulates an initial phase of hyperfiltration before a decline in GFR.28,29 This hypothesis represents a link between our findings and the well-known association between chronic kidney disease and CVD. One factor contributing to hyperfiltration is impaired fasting glucose12 that is associated with manifest CVD.30 However, the association between high GFR and subclinical CVD was independent of glucose in this investigation, suggesting that another mechanism is responsible. Urinary albumin–creatinine ratio (ACR) below the levels representing microalbuminuria predicts both Kidney International (2014) 86, 146–153
clinical investigation
BO Eriksen et al.: Subclinical cardiovascular disease and high GFR
Table 2 | Carotid ultrasonography and left ventricular hypertrophy according to quartile of glomerular filtration rate (the RENIS-T6 study) Quartile of glomerular filtration rate, range (ml/min per 1.73 m2)
Carotid ultrasonography Mean intima media thicknessb, mm Number of atherosclerotic plaques, N (%) 0 1 2 42 Total plaque areac for subjects with plaques, mm2 Left ventricular hypertrophy criteria Sokolow voltage, mV Sokolow voltage X3.5 mV, N (%) Cornell product, mVms Cornell product X244 mVms, N (%) MESA voltage, mV MESA voltage X4.2 mV, N (%)
Quartile 1 (n ¼ 380) (60.9–83.4)
Quartile 2 (n ¼ 380) (83.4–91.8)
Quartile 3 (n ¼ 381) (91.9–101.2)
Quartile 4 (n ¼ 380) (101.2–138.6)
P for a linear trend
0.82 (0.74–0.91)
0.82 (0.74–0.91)
0.82 (0.74–0.90)
0.84 (0.76–0.93)
0.04 0.004
263 (69.1) 85 (22.3) 26 (6.9) 6 (1.7) 15.1 (9.2–23.8)
277 (72.9) 73 (19.2) 25 (6.5) 5 (1.4) 13.3 (9.1–23.9)
272 (71.5) 70 (18.3) 25 (6.5) 14 (3.8) 14.6 (9.1–28.2)
236 (62.0) 87 (23.0) 45 (11.7) 12 (3.3) 18.6 (10.8–30.9)
0.01
2.25 (1.90–2.66) 23 (6.1) 162.5 (130.3–193.1) 25 (6.7) 3.31 (2.79–3.94) 68 (18.0)
2.22 (1.80–2.63) 16 (4.2) 155.0 (123.2–196.8) 26 (6.8) 3.30 (2.69–3.82) 63 (16.6)
2.42 (2.01–2.89) 36 (9.6) 156.6 (119.7–189.2) 30 (8.0) 3.60 (2.96–4.31) 106 (27.7)
2.48 (2.05–2.94) 25 (6.5) 153.4 (108.8–190.2) 31 (8.1) 3.59 (3.09–4.24) 105 (27.6)
o0.0001 0.10 0.01 0.67 o0.0001 o0.0001
Abbreviations: MESA, Multi-Ethnic Study of Atherosclerosis; RENIS-T6, Renal Iohexol Clearance Survey in Tromsø 6. Data are shown as median (interquartile range) or percent. a Quadratic trends across the quartiles were tested for all variables, but none were statistically significant (PX0.05). b Log-transformed when differences between quartiles were tested. c Square-root transformed when differences between quartiles were tested. The median is given for subjects with plaques, but the statistical tests of differences across the quartiles were done for all subjects.
manifest CVD1 and chronic kidney disease.31 However, our findings changed little when urinary ACR was added to the models. The most important strength of this study is that it utilized the largest population-based cohort with GFR measurements, as opposed to inaccurate estimates based on creatinine or cystatin C. The data set includes both carotid ultrasonography and ECG that allowed us to examine subclinical CVD in both the blood vessels and the myocardium. The Tromsø Study is the only study that includes measurements of both IMT and TPA instead of only the presence or absence of plaques. Limitations
Although our results indicate positive linear relationships between GFR, LVH, and carotid atherosclerosis, they are not inconsistent with the well-established association between low GFR and cardiovascular disease. There was a tendency for a U-shaped relationship between GFR and total plaque area (Table 3), but nonlinear models were not statistically significant, possibly because of limited statistical power. Single-sample methods for measuring GFR may not be as precise as multisample methods,32 even though some investigators have found the precision to be very similar.33 Although it is difficult to use the multisample method in population surveys, the possible lower precision of the single-sample method is compensated by a high number of included subjects. As ECG has limited sensitivity for detecting LVH, we may have underestimated the association between high GFR and Kidney International (2014) 86, 146–153
LVH, and we thus may have been unable to detect an association between low GFR and LVH. Ethnicity was not investigated, as the RENIS-T6 included only individuals of European ancestry. Although the interactions between sex and GFR were not statistically significant, statistical power was limited, and that is why we did not present gender-specific analyses. The cross-sectional design of this investigation precludes inferences about causality but adds useful knowledge in the absence of longitudinal studies of subclinical CVD with GFR measurements. Increased carotid IMT,8 carotid plaques,10 and LVH detected by ECG9 are predictors of manifest CVD. Accordingly, our findings indicate that high GFR may be a CVD risk factor. However, longitudinal studies of manifest CVD with measurements of GFR are required for confirmation. Conclusions
We found associations between high GFR and subclinical CVD in the carotid arteries and the heart in apparently healthy individuals that were independent of traditional CVD risk factors, urinary albumin excretion, and fasting glucose. The role of high GFR as a CVD risk factor should be investigated in longitudinal studies. MATERIALS AND METHODS Study population The RENIS-T6 was an ancillary part of the sixth survey of the Tromsø Study,20 which is a series of population-based surveys in the municipality of Tromsø, North Norway. Those invited to the sixth 149
clinical investigation
BO Eriksen et al.: Subclinical cardiovascular disease and high GFR
Table 3 | Multiple logistic regression analyses of carotid atherosclerosis and left ventricular hypertrophy (the RENIS-T6 study) Quartile of glomerular filtration rate as independent variables, range (ml/min per 1.73 m2) Quartile 1 (n ¼ 380) (60.9–83.4) Dependent dichotomous variable
OR
Mean intima media thickness in the upper tertile Model 1 1.00 (Reference) Model 2 1.00 (Reference) Model 3 1.00 (Reference) Model 4 1.00 (Reference)
Quartile 2 (n ¼ 380) (83.4–91.8)
Quartile 3 (n ¼ 381) (91.9–101.2)
OR (95% CI)
OR (95% CI)
Quartile 4 (n ¼ 380) (101.2–138.6)
OR (95% CI)
GFR as continuous independent variable P for linear trend
OR per 10 ml/min per 1.73 m2 increment (95% CI)
(0.81–1.60) (0.84–1.70) (0.84–1.70) (0.83–1.69)
1.11 1.12 1.12 1.11
(0.79–1.58) (0.78–1.60) (0.78–1.60) (0.77–1.59)
1.49 1.45 1.45 1.43
(1.05–2.09)a (1.02–2.07)a (1.02–2.07)a (1.00–2.04)a
0.01 0.02 0.02 0.02
1.14 1.13 1.13 1.12
(1.26–1.04) (1.25–1.02) (1.25–1.02) (1.24–1.02)
Total plaque area greater than the median of non-zero values Model 1 1.00 (Reference) 0.67 (0.42–1.05) Model 2 1.00 (Reference) 0.70 (0.44–1.12) Model 3 1.00 (Reference) 0.70 (0.44–1.13) Model 4 1.00 (Reference) 0.71 (0.45–1.14)
0.86 0.86 0.86 0.88
(0.55–1.35) (0.54–1.35) (0.55–1.37) (0.55–1.39)
1.55 (1.03–2.34)a 1.52 (1.00–2.32)a 1.52 (1.00–2.32) 1.56 (1.02–2.39)a
0.01 0.03 0.03 0.02
1.16 1.14 1.14 1.15
(1.31–1.03) (1.29–1.01) (1.29–1.01) (1.30–1.02)
(1.07–2.27)a (1.05–2.28)a (1.05–2.28)a (1.10–2.38)a
0.05 0.05 0.05 0.03
1.11 1.11 1.12 1.13
(1.00–1.24) (1.00–1.24) (1.00–1.24) (1.01–1.26)
1.14 1.20 1.20 1.19
Left ventricular hypertrophy by either the Sokolow voltage, Cornell product, or MESA voltage criterion Model 1 1.00 (Reference) 1.05 (0.71–1.56) 1.38 (0.95–2.01) Model 2 1.00 (Reference) 1.00 (0.67–1.50) 1.40 (0.95–2.05) Model 3 1.00 (Reference) 1.01 (0.68–1.51) 1.41 (0.96–2.07) Model 4 1.00 (Reference) 1.03 (0.69–1.54) 1.44 (0.98–2.13)
1.56 1.55 1.55 1.62
Abbreviations: CI, confidence interval; GFR, glomerular filtration rate; OR, odds ratio; MESA, Multi-Ethnic Study of Atherosclerosis; RENIS-T6, Renal Iohexol Clearance Survey in Tromsø 6. Model 1 was adjusted for age, sex, body weight, height, body mass index, and the use of lipid-lowering and nonsteroidal anti-inflammatory drugs, angiotensin receptor blockers, angiotensin-converting enzyme inhibitors, b-blockers, calcium blockers, diuretics, and other hypertensive antiinflammatory drugs. Model 2 was adjusted as model 1 and for cardiovascular risk factors (first-degree relative with myocardial infarction before 60 years of age, physical exercise, number of cigarettes currently smoked, ambulatory daytime mean arterial pressure, ambulatory daytime pulse pressure, low-density lipoprotein (LDL) cholesterol, high-density lipoprotein (HDL) cholesterol, fasting triglycerides, and serum high-sensitivity C-reactive protein). Model 3 was adjusted as model 2 and for urinary albumin–creatinine ratio. Model 4 was adjusted as model 3 and for fasting serum glucose. a Po0.05 for difference from quartile 1.
survey (October 2007 to December 2008) included a 40% random sample of individuals aged 50 to 59 years drawn from the population registry and all individuals aged 60 to 62 years (5464 total subjects). A total of 3564 subjects aged 50–62 years completed the survey. We excluded 739 subjects with prevalent myocardial infarction, angina pectoris, stroke, diabetes mellitus, or renal disease. The remaining 2825 eligible subjects were invited to the RENIS-T6. Of the 2107 who responded positively, 12 were excluded because of an allergy to contrast media, iodine, or latex; 65 because of other reasons; and 48 did not present for their appointments. A total of 1982 subjects remained for potential inclusion, and 1632 were investigated according to a predefined target. Five participants were excluded because of technical failures in their GFR measurements, leaving 1627 included participants in the RENIS-T6 cohort (Figure 1). The cohort has previously been shown to be representative of all 2825 eligible subjects.20 For the present analyses, we excluded individuals with chronic kidney disease (GFR o60 ml/min per 1.73 m2; or urinary ACR 41.92 mg/mmol for men and 2.83 mg/mmol for women)34 or diabetes (fasting serum glucose X7.0 mmol/l; or hemoglobinA1C X6.5%).35 A questionnaire about CVD risk factors was administered. Smoking status was divided into current smokers or nonsmokers. The number of cigarettes currently smoked was registered. A family history of early myocardial infarction was defined as a first-degree relative with myocardial infarction before the age of 60 years. Physical exercise was defined as leisure-time activity leading to perspiration or breathlessness. This study complied with the Declaration of Helsinki Principles and was approved by the Norwegian Data Inspectorate 150
and the Regional Committee for Medical and Health Research Ethics of North Norway. All subjects provided written consent. Carotid ultrasonography High-resolution B-mode ultrasonography was performed with a duplex scanner (GE Vivid 7, GE Healthcare, Pittsburgh, PA) with a 12 MHz linear array transducer. The sonographers underwent a 2-month training program before study start, and standard operational procedures were used to minimize errors. Readings were centralized and standardized. The far and near wall of the right common carotid artery, bifurcation, and internal carotid artery (6 locations) were scanned for plaques. A plaque was defined as a localized protrusion with thickening of the vessel wall of 450% compared with the adjacent IMT. Each plaque was outlined on still images with calculation of TPA as the sum of all plaque areas. Automated measurement of the IMT was performed in 10 mm segments on three separate images from the far and near wall of the common carotid artery and the far wall of the bifurcation. The mean IMT from the three images were calculated for each location, and the average of the means in all three locations (mean IMT) was used in the analyses. If present in the predefined location of interest, plaques were included in the IMT measurements. The intra- and interobserver reproducibility were acceptable. Further details have been published previously.36 Electrocardiographic assessment of LVH A standard 12-lead ECG was performed with a Cardiovit AT-104 PC Electrocardiograph by trained medical students and nurses (Schiller AG, Baar, Switzerland). The amplitudes and durations of the different deflections were output automatically by the system. As Kidney International (2014) 86, 146–153
clinical investigation
BO Eriksen et al.: Subclinical cardiovascular disease and high GFR
recommended in current guidelines, more than one diagnostic criterion for LVH was used.37 We applied the Sokolow voltage and the Cornell product,38 as well as the MESA voltage criterion, which has recently been shown to perform well.39 LVH was judged to be present when detected by any of these three criteria. The Sokolow voltage was calculated as the sum of SV1 and the highest of either RV5 or RV6.40 The Cornell product was calculated as the product of the QRS duration and the sum of RaVL and SV3 plus 0.8 for women.38 The MESA voltage was calculated as the sum of SV1, SV2, and RV5.39 The thresholds indicating LVH were X3.5 mV, X244 mV ms, and X4.2 mV, respectively. Iohexol clearance GFR was measured as the single-sample plasma clearance of iohexol. This method has been validated against gold standard methods. Five milliliters of iohexol (Omnipaque, 300 mg I/ml, Amersham Health, London, UK) was injected intravenously. The exact time was measured to sampling for measurement of iohexol concentration by high-performance liquid chromatography. Details have been published previously.20 Blood pressure Ambulatory BP was measured with the Spacelab 90207 (Spacelab, Redmond, WA). The daytime mean systolic and diastolic ambulatory BPs were calculated as the weighted mean of the measurements from 10.00 to 20.00. Mean pulse pressure was defined as the mean systolic minus the mean diastolic ambulatory BP. Mean arterial pressure was defined as the mean diastolic ambulatory BP plus one-third of the pulse pressure. Individuals with conventional systolic BP 4140 mm Hg, conventional diastolic BP 490 mm Hg, or individuals using antihypertensive medication were categorized as having hypertension. The methods have been described previously.11 Other measurements Fasting serum glucose, triglycerides, and low-density lipoprotein and high-density lipoprotein cholesterol were measured on the Modular model P800 (Roche Diagnostics, Indianapolis, IN). Hemoglobin A1C was measured with a liquid chromatographic method (Variant II instrument, Bio-Rad Laboratories, Hercules, CA). Three consecutive samples of first-void spot urine were analyzed for albumin and creatinine as described previously.41 The ACR was calculated for each specimen, and the median ACR was used in the analyses. Statistical analysis Continuous variables are presented as the median (interquartile range). Differences across GFR quartiles were tested with analysis of variance, median, ordinal logistic, logistic, or Poisson regression as appropriate. Tests for unadjusted linear and quadratic trends were performed using linear and squared terms for GFR in the regressions. Serum triglycerides and mean IMT were logtransformed, and the TPA was square root-transformed to reduce skewness. Adjusted models of mean IMT, TPA, and LVH were analyzed separately with multiple logistic regressions using GFR as the independent variable. For the mean IMT, the dependent dichotomous variable was defined as the upper tertile. Because the majority of the subjects did not have atherosclerotic plaques, the dependent dichotomous variable of TPA was defined as TPA greater Kidney International (2014) 86, 146–153
than the median of non-zero values. The dichotomous LVH variable was defined as LVH detected by the Sokolow voltage, Cornell product, or MESA voltage criteria. The three criteria were also analyzed as the dependent continuous variables in multiple linear regression analyses. Four subjects with left bundle branch block were excluded from the LVH analyses.37 For each of the dependent variables above, separate multiple logistic regression analyses with GFR as an independent variable both divided into quartiles and as a continuous variable were performed. We adjusted for age, gender, body weight, height, body mass index, and the use of lipid-lowering and nonsteroidal anti-inflammatory drugs, angiotensin receptor blockers, angiotensin-converting enzyme inhibitors, b-blockers, calcium blockers, diuretics, and other hypertensives as dichotomous variables (model 1). Model 2 used the same factors as model 1 as well as a family history of early myocardial infarction (yes/no), physical exercise (yes/no), daytime ambulatory mean arterial pressure and pulse pressure, number of currently smoked cigarettes, low-density lipoprotein and high-density lipoprotein cholesterol, the log of fasting triglycerides, and high-sensitivity C-reactive protein. Model 3 used the same factors as model 2 as well as urinary ACR. Model 4 used the same factors as model 3 as well as fasting glucose. Body mass index was calculated as body weight divided by the square of height. Interactions between sex and GFR were tested in all fully adjusted models. Carotid data were missing for 3.6% of the cohort, and ECG data for 16.4%. In accordance with current recommendations, missing data were imputed using multiple chained imputation in STATA/MP 12.1 (StataCorp LP, College Station, TX, www.stata.com; see Supplementary Methods online, including Supplementary Table S1 and Supplementary Figures S1–S6).42 All statistical tests were performed on 50 imputed data sets using mi estimate in STATA, except for the nonlinear analyses described below, which are not supported by mi estimate and were performed for complete cases. Nonlinear associations between the dependent variables and GFR or any of the adjustment variables in the fully adjusted models were analyzed in multivariable second-degree fractional polynomial models with mfp in STATA. GFR was analyzed in the customary way as standardized to 1.73 m2 of body surface area. As this is a disputed issue, all the analyses above were repeated with GFR standardized to 40 l of estimated total body water (GFR40), as suggested by us in a previous publication.20 We also repeated the analyses with absolute (unadjusted) GFR measured in ml/min. Statistical significance was defined as Po0.05, and all statistical tests were two sided. DISCLOSURE
All the authors declared no competing interests. ACKNOWLEDGMENTS
We are deeply grateful for the valuable advice from the late Professor Egil Arnesen, former principal investigator of the Tromsø Study, who was active with this study until his death in December 2009. We thank Bjørg Skog Høgset, Saskia van Heusden, and the staff at the Clinical Research Unit (University Hospital of North Norway) for assistance in planning the study, performing the procedures, and collecting data according to the GCP standard. We also thank Harald Strand and the staff at the Department of Medical Biochemistry (University Hospital of North Norway) for HPLC analyses of iohexol. We also want to thank all the participants in the RENIS-T6 cohort for making this study possible. 151
clinical investigation
SUPPLEMENTARY MATERIAL Table S1. Missing data. Table S2. Multiple logistic regression analyses of electrocardiographic criteria of left ventricular hypertrophy by quartiles of GFR, and by GFR as a continuous variable. Table S3. Multiple linear regression analyses of electrocardiographic criteria of left ventricular hypertrophy by quartiles of GFR, and by GFR as a continuous variable. Table S4. Study population characteristics according to tertile of mean carotid intima-media thickness. Table S5. Study population characteristics according to category of carotid total plaque area. Table S6. Study population characteristics according to left ventricular hypertrophy. Figure S1. Comparison of the cumulative distribution functions of observed and imputed Sokolow voltage for every tenth imputation. Figure S2. Comparison of the cumulative distribution functions of observed and imputed Cornell product for every tenth imputation. Figure S3. Comparison of the cumulative distribution functions of observed and imputed MESA voltage for every tenth imputation. Figure S4. Comparison of the cumulative distribution functions of observed and imputed mean intima-media thickness for every tenth imputation. Figure S5. Comparison of the cumulative distribution functions of observed and imputed total plaque area for every tenth imputation. Figure S6. Comparison of the histograms of observed and imputed number of carotid plaques for every tenth imputation. Supplementary material is linked to the online version of the paper at http://www.nature.com/ki
BO Eriksen et al.: Subclinical cardiovascular disease and high GFR
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23. 24.
REFERENCES Chronic Kidney Disease Prognosis Consortium. Association of estimated glomerular filtration rate and albuminuria with all-cause and cardiovascular mortality in general population cohorts: a collaborative meta-analysis. Lancet 2010; 375: 2073–2081. 2. Nitsch D, Grams M, Sang Y et al. Associations of estimated glomerular filtration rate and albuminuria with mortality and renal failure by sex: a meta-analysis. BMJ 2013; 346: f324. 3. Di Angelantonio E, Chowdhury R, Sarwar N et al. Chronic kidney disease and risk of major cardiovascular disease and non-vascular mortality: prospective population based cohort study. BMJ 2010; 341: c4986. 4. Inker LA, Schmid CH, Tighiouart H et al. Estimating glomerular filtration rate from serum creatinine and cystatin C. N Engl J Med 2012; 367: 20–29. 5. Mathisen UD, Melsom T, Ingebretsen OC et al. Estimated GFR is associated with cardiovascular risk factors independently of measured GFR. J Am Soc Nephrol 2011; 22: 927–937. 6. Knight EL, Verhave JC, Spiegelman D et al. Factors influencing serum cystatin C levels other than renal function and the impact on renal function measurement. Kidney Int 2004; 65: 1416–1421. 7. Rule AD, Bailey KR, Lieske JC et al. Estimating the glomerular filtration rate from serum creatinine is better than from cystatin C for evaluating risk factors associated with chronic kidney disease. Kidney Int 2013; 83: 1169–1176. 8. Den Ruijter HM, Peters SE, Anderson TJ et al. Common carotid intimamedia thickness measurements in cardiovascular risk prediction: a metaanalysis. JAMA 2012; 308: 796–803. 9. Chou R, Arora B, Dana T et al. Screening asymptomatic adults with resting or exercise electrocardiography: a review of the evidence for the U.S. Preventive Services Task Force. Ann Intern Med 2011; 155: 375–385. 10. Wyman RA, Mays ME, McBride PE et al. Ultrasound-detected carotid plaque as a predictor of cardiovascular events. Vasc Med 2006; 11: 123–130. 11. Mathisen UD, Melsom T, Ingebretsen OC et al. Ambulatory blood pressure is associated with measured glomerular filtration rate in the general middle-aged population. J Hypertens 2012; 30: 497–504. 12. Melsom T, Mathisen UD, Ingebretsen OC et al. Impaired fasting glucose is associated with renal hyperfiltration in the general population. Diabetes Care 2011; 34: 1546–1551.
25.
1.
152
26.
27.
28. 29. 30.
31.
32.
33.
34.
35. 36.
37.
Bui AL, Katz R, Kestenbaum B et al. Cystatin C and carotid intima-media thickness in asymptomatic adults: the Multi-Ethnic Study of Atherosclerosis (MESA). Am J Kidney Dis 2009; 53: 389–398. Bobbert T, Mai K, Fischer-Rosinsky A et al. Relation between physiological variation of renal function and carotid intima media thickness in nondiabetic individuals. J Atheroscler Thromb 2010; 17: 242–248. Desbien AM, Chonchol M, Gnahn H et al. Kidney function and progression of carotid intima-media thickness in a community study. Am J Kidney Dis 2008; 51: 584–593. Kastarinen H, Ukkola O, Kesaniemi YA. Glomerular filtration rate is related to carotid intima-media thickness in middle-aged adults. Nephrol Dial Transplant 2009; 24: 2767–2772. Choi SW, Kim HY, Lee YH et al. eGFR is associated with subclinical atherosclerosis independent of albuminuria: the Dong-gu Study. Atherosclerosis 2010; 212: 661–667. Jiang SM, Sun XF, Gu HX et al. Effects of decline in renal function with age on the outcome of asymptomatic carotid plaque in healthy adults: a 5-year follow-up study. Chin Med J 2012; 125: 2649–2657. Moran A, Katz R, Jenny NS et al. Left ventricular hypertrophy in mild and moderate reduction in kidney function determined using cardiac magnetic resonance imaging and cystatin C: the Multi-Ethnic Study of Atherosclerosis (MESA). Am J Kidney Dis 2008; 52: 839–848. Eriksen BO, Melsom T, Mathisen UD et al. GFR normalized to total body water allows comparisons across genders and body sizes. J Am Soc Nephrol 2011; 22: 1517–1525. Delanaye P, Radermecker RP, Rorive M et al. Indexing glomerular filtration rate for body surface area in obese patients is misleading: concept and example. Nephrol Dial Transplant 2005; 20: 2024–2028. Okada R, Yasuda Y, Tsushita K et al. Glomerular hyperfiltration in prediabetes and prehypertension. Nephrol Dial Transplant 2011; 27: 1821–1825. Tomaszewski M, Charchar FJ, Maric C et al. Glomerular hyperfiltration: a new marker of metabolic risk. Kidney Int 2007; 71: 816–821. Maeda I, Hayashi T, Sato KK et al. Cigarette smoking and the association with glomerular hyperfiltration and proteinuria in healthy middle-aged men. Clin J Am Soc Nephrol 2011; 6: 2462–2469. Pinto-Sietsma SJ, Mulder J, Janssen WM et al. Smoking is related to albuminuria and abnormal renal function in nondiabetic persons. Ann Intern Med 2000; 133: 585–591. Verhave JC, Hillege HL, Burgerhof JG et al. The association between atherosclerotic risk factors and renal function in the general population. Kidney Int 2005; 67: 1967–1973. Licht CM, de Geus EJ, Penninx BW. Dysregulation of the autonomic nervous system predicts the development of the metabolic syndrome. J Clin Endocrinol Metab 2013; 98: 2484–2493. Brenner BM, Lawler EV, Mackenzie HS. The hyperfiltration theory: a paradigm shift in nephrology. Kidney Int 1996; 49: 1774–1777. Ruggenenti P, Remuzzi G. Time to abandon microalbuminuria? Kidney Int 2006; 70: 1214–1222. Ford ES, Zhao GX, Li CY. Pre-diabetes and the risk for cardiovascular disease: a systematic review of the evidence. J Am Coll Cardiol 2010; 55: 1310–1317. Hallan SI, Ritz E, Lydersen S et al. Combining GFR and albuminuria to classify CKD improves prediction of ESRD. J Am Soc Nephrol 2009; 20: 1069–1077. Gaspari F, Guerini E, Perico N et al. Glomerular filtration rate determined from a single plasma sample after intravenous iohexol injection: is it reliable? J Am Soc Nephrol 1996; 7: 2689–2693. Bird NJ, Peters C, Michell AR et al. Comparison of GFR measurements assessed from single versus multiple samples. Am J Kidney Dis 2009; 54: 278–288. K/DOQI clinical practice guidelines for chronic kidney disease work group. Part 4. Definition and classification of stages of chronic kidney disease. Am J Kidney Dis 2002; 39: S46–S75. American Diabetes Association. Diagnosis and classification of diabetes mellitus. Diabetes Care 2011; 34: S62–S69. Herder M, Johnsen SH, Arntzen KA et al. Risk factors for progression of carotid intima-media thickness and total plaque area: a 13-year follow-up study: the Tromso Study. Stroke 2012; 43: 1818–1823. Hancock EW, Deal BJ, Mirvis DM et al. AHA/ACCF/HRS recommendations for the standardization and interpretation of the electrocardiogram: part V: electrocardiogram changes associated with cardiac chamber hypertrophy: a scientific statement from the American Heart Association Electrocardiography and Arrhythmias Committee, Council on Clinical Cardiology; the American College of Cardiology Foundation; and the Heart Rhythm Society Endorsed by the International Kidney International (2014) 86, 146–153
clinical investigation
BO Eriksen et al.: Subclinical cardiovascular disease and high GFR
38.
39.
Society for Computerized Electrocardiology. J Am Coll Cardiol 2009; 53: 992–1002. Devereux RB, Bella J, Boman K et al. Echocardiographic left ventricular geometry in hypertensive patients with electrocardiographic left ventricular hypertrophy: the LIFE study. Blood Pressure 2001; 10: 74–82. Jain A, Tandri H, Dalal D et al. Diagnostic and prognostic utility of electrocardiography for left ventricular hypertrophy defined by magnetic resonance imaging in relationship to ethnicity: the Multi-Ethnic Study of Atherosclerosis (MESA). Am Heart J 2010; 159: 652–658.
Kidney International (2014) 86, 146–153
40.
41.
42.
Sokolow M, Lyon TP. The ventricular complex in left ventricular hypertrophy as obtained by unipolar precordial and limb leads. Am Heart J 1949; 37: 161–186. Solbu MD, Kronborg J, Eriksen BO et al. Cardiovascular risk-factors predict progression of urinary albumin-excretion in a general, non-diabetic population: a gender-specific follow-up study. Atherosclerosis 2008; 201: 398–406. Little RJ, D’Agostino R, Cohen ML et al. The prevention and treatment of missing data in clinical trials. N Engl J Med 2012; 367: 1355–1360.
153
http://www.kidney-international.org & 2013 International Society of Nephrology
see commentary on page 10
Subclinical cardiovascular disease is associated with a high glomerular filtration rate in the nondiabetic general population Bjørn O. Eriksen1,2,3, Maja-Lisa Løchen4, Kjell A. Arntzen1,5, Geir Bertelsen1,6, Britt-Ann W. Eilertsen3, Therese von Hanno1,7, Marit Herder4,8, Trond G. Jenssen1,9, Ulla D. Mathisen2, Toralf Melsom2, Inger Njølstad4, Marit D. Solbu2, Ingrid Toft1,2 and Ellisiv B. Mathiesen1,5 1
Department of Clinical Medicine, Faculty of Health Sciences, University of Tromsø, Tromsø, Norway; 2Section of Nephrology, University Hospital of North Norway, Tromsø, Norway; 3Department of Clinical Research, University Hospital of North Norway, Tromsø, Norway; 4 Department of Community Medicine, Faculty of Health Sciences, University of Tromsø, Tromsø, Norway; 5Department of Neurology and Neurophysiology, University Hospital of North Norway, Tromsø, Norway; 6Department of Ophthalmology, University Hospital of North Norway, Tromsø, Norway; 7Department of Ophthalmology, Nordland Hospital, Bodø, Norway; 8Department of Radiology, University Hospital of North Norway, Tromsø, Norway and 9Department of Nephrology, Oslo University Hospital, Oslo, Norway
A reduced glomerular filtration rate (GFR) in chronic kidney disease is a risk factor for cardiovascular disease. However, evidence indicates that a high GFR may also be a cardiovascular risk factor. This issue remains unresolved due to a lack of longitudinal studies of manifest cardiovascular disease with precise GFR measurements. Here, we performed a cross-sectional study of the relationship between high GFR measured as iohexol clearance and subclinical cardiovascular disease in the Renal Iohexol Clearance Survey in Tromsø 6 (RENIS-T6), a representative sample of the middle-aged general population. A total of 1521 persons without cardiovascular disease, chronic kidney disease, diabetes, or micro- or macroalbuminuria were examined with carotid ultrasonography and electrocardiography. The GFR in the highest quartile was associated with an increased odds ratio of having total carotid plaque area greater than the median of non-zero values (odds ratio 1.56, 95% confidence interval 1.02–2.39) or electrocardiographic signs of left ventricular hypertrophy (odds ratio 1.62, 95% confidence interval 1.10–2.38) compared to the lowest quartile. The analyses were adjusted for cardiovascular risk factors, urinary albumin excretion, and fasting serum glucose. Thus, high GFR is associated with carotid atherosclerosis and left ventricular hypertrophy and should be investigated as a possible risk factor for manifest cardiovascular disease in longitudinal studies. Kidney International (2014) 86, 146–153; doi:10.1038/ki.2013.470; published online 4 December 2013 KEYWORDS: arteriosclerosis; cardiovascular disease; glomerular filtration rate; hyperfiltration; left ventricular hypertrophy; risk factors Correspondence: Bjørn O. Eriksen, Section of Nephrology, University Hospital of North Norway, 9038 Tromsø, Norway. E-mail: [email protected] Received 7 June 2013; revised 20 August 2013; accepted 19 September 2013; published online 4 December 2013 146
Low glomerular filtration rate (GFR) is a risk factor for cardiovascular disease (CVD).1 Although high GFR has been regarded as normal, evidence suggests that it may also confer increased risk. In a pooled analysis of over two million subjects, Nitsch et al.2 found an increased cardiovascular mortality for nonproteinuric men with high GFR, whereas a lower increase for women did not reach statistical significance. These analyses were adjusted for the presence of diabetes. Another study found an increased risk of coronary heart disease with high GFR in individuals without prevalent CVD, diabetes, or chronic kidney disease.3 As these studies used serum creatinine to estimate GFR, it has been assumed that these and similar findings can be explained by an association between chronic disease and low muscle mass that would result in overestimated GFR. However, this assumption has not been investigated. Because precise GFR measurements are costly and cumbersome, there are, to our knowledge, no longitudinal population-based studies of measured GFR and CVD. GFR estimates based on creatinine or cystatin C are known to be inaccurate in the upper range.4 In addition, creatinine and cystatin C are influenced by non-GFR factors, including cardiovascular risk factors, that may cause confounding.5–7 Accordingly, whether high GFR is a CVD risk factor remains unresolved. However, the hypothesis implies that subclinical CVD may be cross-sectionally associated with high GFR. We investigated this issue in the Renal Iohexol Clearance Survey in Tromsø 6 (RENIS-T6) that has the largest population-based cohort with GFR measurements. The cohort consists of a representative sample of the middle-aged general population without manifest CVD, diabetes, or chronic kidney disease, and who underwent carotid ultrasonography and electrocardiography (ECG). Carotid atherosclerosis and left ventricular hypertrophy (LVH) are signs of subclinical CVD that predict manifest CVD.8–10 Our aim was to investigate whether Kidney International (2014) 86, 146–153
clinical investigation
BO Eriksen et al.: Subclinical cardiovascular disease and high GFR
there is a cross-sectional association between subclinical CVD and high GFR in apparently healthy individuals. RESULTS
Of the 1627 subjects included in the RENIS-T6, individuals with GFR o60 ml/min per 1.73 m2 (n ¼ 34), micro- or macroalbuminuria (n ¼ 44), or diabetes (n ¼ 33) were excluded. Three individuals had both diabetes and microalbuminuria; two had both diabetes and GFR o60 ml/min per 1.73 m2. Accordingly, 106 subjects were excluded, leaving 1521 for the present analyses (Figure 1). There were differences across the GFR quartiles with regard to age; gender; height; body weight; current smoking; ambulatory systolic blood pressure (BP), diastolic BP, mean arterial pressure, and pulse pressure; antihypertensive drug use; high-density lipoprotein cholesterol and fasting glucose (Po0.05) (Table 1). These findings have been discussed in previous publications.5,11,12 In unadjusted analyses, higher GFR quartile was associated with greater carotid mean intima media thickness (IMT), a higher number of carotid plaques, greater total plaque area (TPA), greater Sokolow and Multi-Ethnic Study of Atherosclerosis (MESA) voltage, and a higher percentage of LVH detected by the MESA voltage criterion (Po0.05 for N = 5464 Invited to the Tromsø 6 Study and aged 50 to 62 years
N = 3564 Met and completed the main Tromsø 6 Study
N = 2825 Invited to RENIS-T6
N = 2107 Responders
N = 739 Reported a previous myocardial infarction, angina pectoris, stroke, diabetes mellitus, or any renal disease except urinary tract infection.
N = 125 Excluded because of allergy to contrast media, iodine, or latex, or for other reasons. Includes 48 who withdrew.
N = 1982 Eligible for inclusion
N = 1632 Investigated in RENIS-T6 according to a predetermined target
N = 1627 The RENIS-T6 cohort
N = 1521 Present study population
N=5 Technical failure in the iohexol-clearance measurements. N = 106 Diabetes (N = 33) according to their fasting plasma glucose or HbA1c. CKD (N = 34) defined as measured GFR< 60 ml/min/m or micro- or macroalbuminuria (N = 44) (3 persons had both diabetes and microalbuminuria, 2 both diabetes and CKD)
Figure 1 | Study population derived from RENIS-T6, an ancillary investigation of the sixth Tromsø Study. CKD, chronic kidney disease; GFR, glomerular filtration rate; RENIS-T6, Renal Iohexol Clearance Survey in Tromsø 6. Kidney International (2014) 86, 146–153
linear trend; Table 2). There was an opposite trend for the Cornell product (Table 2). In the calculation of the Cornell product, a factor of 0.8 is added for women, who have a lower GFR than men, which explains this finding. Quadratic trends were not statistically significant for any of the variables in Table 2. More information about the relationships between covariates and the outcomes can be found in Supplementary Tables S4 to S6 online. In the adjusted models (Table 3), the odds ratios (ORs) for the dichotomous IMT and TPA variables increased with increasing quartile of GFR (Po0.05). The ORs for the highest quartile compared with the lowest quartile were 1.49 for IMT and 1.56 for TPA in the fully adjusted models (model 4; Po0.05). Both IMT and TPA were also associated with GFR as a continuous variable (Po0.05). In the adjusted models, LVH was judged to be present when detected by either the Sokolow or MESA voltage criterion, or the Cornell product criterion. In all the adjusted models (Table 3), the OR for LVH was elevated for the highest compared with the lowest GFR quartile (Po0.05). For the fully adjusted model, the OR was 1.62. When LVH by each criterion was analyzed separately, OR was increased in the highest GFR quartile in all the adjusted models for the Cornell product criterion (Po0.05), and there was a linear trend for increasing OR with increasing GFR for the MESA voltage in the fully adjusted model (Po0.05)(Supplementary Table S2 online). LVH as detected by the Sokolow voltage criterion was not associated with GFR. There were positive associations between both the Sokolow voltage and MESA voltage analyzed as continuous variables and GFR (Po0.05), but there was no association with the Cornell product (Supplementary Table S3 online). None of the interactions between sex and GFR in any of the fully adjusted models in Table 3 were statistically significant. When analyzing fractional polynomial models with the same dependent variables as in Table 3 and the same independent variables as in the fully adjusted models (model 4), no nonlinear effects of GFR were found (PX0.05). To explore possible confounding by body size, all the multiple regression analyses were repeated with GFR standardized to a total body water of 40 l (GFR40) as the dependent variable instead of GFR standardized to body surface area. The results of these analyses were similar, as were also the results of analyses using absolute GFR in ml/min (not shown). DISCUSSION
To our knowledge, this investigation is the first to find an independent association between high normal GFR and subclinical CVD manifested as carotid atherosclerosis and LVH in the healthy general population. The cross-sectional relationship between GFR and carotid atherosclerosis has been examined previously. Several studies have found no statistically significant association between GFR and IMT,13,14 whereas two studies found an association 147
clinical investigation
BO Eriksen et al.: Subclinical cardiovascular disease and high GFR
Table 1 | Study population characteristics according to quartile of glomerular filtration rate (the RENIS-T6 study) Quartile of glomerular filtration rate, range (ml/min per 1.73 m2)
Glomerular filtration rate, ml/min per 1.73 m2 Female gender, N (%) Age, years Height, cm Body weight, kg Body mass index, kg/m2 First-degree relative with myocardial infarction before 60 years, N (%) Physical exercisea, N (%) Current smoking, N (%) Mean no. of cigarettes smoked per day, if currently smoking LDL cholesterol, mmol/l HDL cholesterol, mmol/l Triglyceridesb, mmol/l High-sensitivity C–reactive protein, mg/l Fasting serum glucose, mmol/l Urine albumin–creatinine ratio, mg/mmol
Quartile 1 (n ¼ 380) (60.9–83.4)
Quartile 2 (n ¼ 380) (83.4–91.8)
Quartile 3 (n ¼ 381) (91.9–101.2)
Quartile 4 (n ¼ 380) (101.2–138.6)
P for linear trend
77.6 (72.8–80.8) 256 (67.4) 60.5 (56.6–62) 169.2 (163.1–175.5) 77.6 (68.4–88.6) 26.7 (24.4–29.7) 88 (23.2)
87.6 (85.4–89.7) 211 (55.5) 59.2 (54.7–61.5) 169.7 (164.0–177.0) 77.5 (68.1–87.1) 26.5 (24.2–29.3) 62 (16.3)
96.1 (94.0–98.7) 184 (48.3) 57.8 (54.2–61.) 169.8 (164.0–177.2) 77.9 (69.4–88.9) 26.9 (24.5–29.5) 80 (21.0)
107.6 (103.9–113.7) 120 (31.6) 57.1 (53.6–60.8) 173.0 (166.4–178.2) 81.4 (71.2–90.7) 27.0 (24.7–29.9) 72 (18.9)
o0.0001 o0.0001 o0.0001 0.02 0.95 o0.0001
195 (51.3) 73 (19.3) 10.4
228 (60.1) 63 (16.6) 11.5
228 (59.9) 73 (19.2) 12.1
197 (51.8) 111 (29.2) 11.8
0.94d o0.0001d o0.0001d
3.60 1.50 1.00 1.21 5.20 0.29
(3.00–4.20) (1.30–1.85) (0.80–1.50) (0.66–2.21) (4.90–5.50) (0.17–0.54)
3.60 1.50 1.00 1.15 5.30 0.28
(3.10–4.20) (1.30–1.80) (0.70–1.30) (0.61–2.16) (5.00–5.50) (0.16–0.52)
3.70 1.50 1.10 1.17 5.30 0.26
(3.20–4.30) (1.20–1.80) (0.70–1.50) (0.65–2.16) (5.00–5.60) (0.15–0.46)
3.60 1.40 1.10 1.18 5.30 0.29
(3.10–4.20) (1.20–1.70) (0.70–1.50) (0.64–2.08) (5.10–5.80) (0.15–0.54)
0.45 o0.0001 0.13d 0.76 o0.0001 0.63d
Daytime ambulatory blood pressure Systolic, mm Hg Diastolic, mm Hg Mean arterial, mm Hg Pulse pressure, mm Hg
127.8 (119.2–137.3) 80.5 (74.3–86.3) 97.2 (90.5–103.9) 47.0 (42.3–52.7)
128.3 (119.3–136.1) 81.5 (75.8–86.4) 98.1 (91.1–103.6) 46.3 (41.6–51.2)
129.1 (120.4–137.6) 81.9 (76.2–86.9) 98.3 (91.6–104.) 46.5 (42.2–51.9)
132.1 (123.5–140.1) 83.3 (77.7–88.3) 99.6 (93.8–105.6) 48.2 (43.1–53.0)
o0.0001 o0.0001 o0.0001 0.04
Hypertensionc, N (%) Antihypertensive medication, N (%)
171 (45.0) 81 (21.3)
160 (42.1) 62 (16.3)
159 (41.7) 64 (16.8)
145 (38.2) 58 (15.3)
0.06 0.01
Abbreviations: HDL, high-density lipoprotein; LDL, low-density lipoprotein; RENIS-T6, Renal Iohexol Clearance Survey in Tromsø 6. Data are shown as median (interquartile range) or percent. a Defined as leisure-time physical exercise leading to perspiration or breathlessness (yes/no). b Log-transformed when differences between quartiles were tested. c Defined as conventional systolic blood pressure (BP) 4140, conventional diastolic BP 490, or the use of antihypertensive medication. d Po0.05 for quadratic trend.
between low GFR and increased IMT.15,16 The conflicting results may reflect shortcomings of the GFR assessments, as all these studies used GFR estimates rather than measured GFR. Two Asian studies of GFR and plaques also showed differing results.17,18 These studies investigated only the presence or absence of plaques, not plaque area. To our knowledge, there are no population-based studies of the relationship between GFR and LVH based on ECG findings. However, MESA investigated GFR and LV mass measured with magnetic resonance imaging at baseline.19 For a cystatin C–based estimated GFR o75 ml/min per 1.73 m2, both the ORs for LVH and LV mass were increased, but a linear regression found no association for estimated GFR 475 ml/min per 1.73 m2. Because this analysis was unadjusted, it does not exclude an association between high GFR and LVH. In addition, cystatin C–based estimated GFR is known to be influenced by several non-GFR factors that may have affected the result.5–7 LVH detected by ECG is known to be confounded by body mass index. For this reason, we performed analyses in which GFR was standardized to total body water, which avoids confounding from body size.20 In addition, we repeated the analyses with absolute GFR in ml/min, which have been used by some authors to avoid underestimation of renal function 148
in obese individuals.21 The results of these analyses differed little from the analyses with GFR standardized to body surface area that indicates that body size is not a confounder of the association between LVH and GFR. The association between carotid atherosclerosis, LVH, and high normal GFR may seem paradoxical. However, associations between increased GFR and prediabetes, high blood pressure, increased pulse rate, smoking, and other cardiovascular risk factors have been found in RENIS-T6 and other studies of nondiabetic subjects.5,11,12,22–26 The mechanism for these associations has not been established, but increased activity of the sympathetic nervous system may be one contributing factor.27 In addition, one hypothesis regarding the pathogenesis of chronic kidney disease postulates an initial phase of hyperfiltration before a decline in GFR.28,29 This hypothesis represents a link between our findings and the well-known association between chronic kidney disease and CVD. One factor contributing to hyperfiltration is impaired fasting glucose12 that is associated with manifest CVD.30 However, the association between high GFR and subclinical CVD was independent of glucose in this investigation, suggesting that another mechanism is responsible. Urinary albumin–creatinine ratio (ACR) below the levels representing microalbuminuria predicts both Kidney International (2014) 86, 146–153
clinical investigation
BO Eriksen et al.: Subclinical cardiovascular disease and high GFR
Table 2 | Carotid ultrasonography and left ventricular hypertrophy according to quartile of glomerular filtration rate (the RENIS-T6 study) Quartile of glomerular filtration rate, range (ml/min per 1.73 m2)
Carotid ultrasonography Mean intima media thicknessb, mm Number of atherosclerotic plaques, N (%) 0 1 2 42 Total plaque areac for subjects with plaques, mm2 Left ventricular hypertrophy criteria Sokolow voltage, mV Sokolow voltage X3.5 mV, N (%) Cornell product, mVms Cornell product X244 mVms, N (%) MESA voltage, mV MESA voltage X4.2 mV, N (%)
Quartile 1 (n ¼ 380) (60.9–83.4)
Quartile 2 (n ¼ 380) (83.4–91.8)
Quartile 3 (n ¼ 381) (91.9–101.2)
Quartile 4 (n ¼ 380) (101.2–138.6)
P for a linear trend
0.82 (0.74–0.91)
0.82 (0.74–0.91)
0.82 (0.74–0.90)
0.84 (0.76–0.93)
0.04 0.004
263 (69.1) 85 (22.3) 26 (6.9) 6 (1.7) 15.1 (9.2–23.8)
277 (72.9) 73 (19.2) 25 (6.5) 5 (1.4) 13.3 (9.1–23.9)
272 (71.5) 70 (18.3) 25 (6.5) 14 (3.8) 14.6 (9.1–28.2)
236 (62.0) 87 (23.0) 45 (11.7) 12 (3.3) 18.6 (10.8–30.9)
0.01
2.25 (1.90–2.66) 23 (6.1) 162.5 (130.3–193.1) 25 (6.7) 3.31 (2.79–3.94) 68 (18.0)
2.22 (1.80–2.63) 16 (4.2) 155.0 (123.2–196.8) 26 (6.8) 3.30 (2.69–3.82) 63 (16.6)
2.42 (2.01–2.89) 36 (9.6) 156.6 (119.7–189.2) 30 (8.0) 3.60 (2.96–4.31) 106 (27.7)
2.48 (2.05–2.94) 25 (6.5) 153.4 (108.8–190.2) 31 (8.1) 3.59 (3.09–4.24) 105 (27.6)
o0.0001 0.10 0.01 0.67 o0.0001 o0.0001
Abbreviations: MESA, Multi-Ethnic Study of Atherosclerosis; RENIS-T6, Renal Iohexol Clearance Survey in Tromsø 6. Data are shown as median (interquartile range) or percent. a Quadratic trends across the quartiles were tested for all variables, but none were statistically significant (PX0.05). b Log-transformed when differences between quartiles were tested. c Square-root transformed when differences between quartiles were tested. The median is given for subjects with plaques, but the statistical tests of differences across the quartiles were done for all subjects.
manifest CVD1 and chronic kidney disease.31 However, our findings changed little when urinary ACR was added to the models. The most important strength of this study is that it utilized the largest population-based cohort with GFR measurements, as opposed to inaccurate estimates based on creatinine or cystatin C. The data set includes both carotid ultrasonography and ECG that allowed us to examine subclinical CVD in both the blood vessels and the myocardium. The Tromsø Study is the only study that includes measurements of both IMT and TPA instead of only the presence or absence of plaques. Limitations
Although our results indicate positive linear relationships between GFR, LVH, and carotid atherosclerosis, they are not inconsistent with the well-established association between low GFR and cardiovascular disease. There was a tendency for a U-shaped relationship between GFR and total plaque area (Table 3), but nonlinear models were not statistically significant, possibly because of limited statistical power. Single-sample methods for measuring GFR may not be as precise as multisample methods,32 even though some investigators have found the precision to be very similar.33 Although it is difficult to use the multisample method in population surveys, the possible lower precision of the single-sample method is compensated by a high number of included subjects. As ECG has limited sensitivity for detecting LVH, we may have underestimated the association between high GFR and Kidney International (2014) 86, 146–153
LVH, and we thus may have been unable to detect an association between low GFR and LVH. Ethnicity was not investigated, as the RENIS-T6 included only individuals of European ancestry. Although the interactions between sex and GFR were not statistically significant, statistical power was limited, and that is why we did not present gender-specific analyses. The cross-sectional design of this investigation precludes inferences about causality but adds useful knowledge in the absence of longitudinal studies of subclinical CVD with GFR measurements. Increased carotid IMT,8 carotid plaques,10 and LVH detected by ECG9 are predictors of manifest CVD. Accordingly, our findings indicate that high GFR may be a CVD risk factor. However, longitudinal studies of manifest CVD with measurements of GFR are required for confirmation. Conclusions
We found associations between high GFR and subclinical CVD in the carotid arteries and the heart in apparently healthy individuals that were independent of traditional CVD risk factors, urinary albumin excretion, and fasting glucose. The role of high GFR as a CVD risk factor should be investigated in longitudinal studies. MATERIALS AND METHODS Study population The RENIS-T6 was an ancillary part of the sixth survey of the Tromsø Study,20 which is a series of population-based surveys in the municipality of Tromsø, North Norway. Those invited to the sixth 149
clinical investigation
BO Eriksen et al.: Subclinical cardiovascular disease and high GFR
Table 3 | Multiple logistic regression analyses of carotid atherosclerosis and left ventricular hypertrophy (the RENIS-T6 study) Quartile of glomerular filtration rate as independent variables, range (ml/min per 1.73 m2) Quartile 1 (n ¼ 380) (60.9–83.4) Dependent dichotomous variable
OR
Mean intima media thickness in the upper tertile Model 1 1.00 (Reference) Model 2 1.00 (Reference) Model 3 1.00 (Reference) Model 4 1.00 (Reference)
Quartile 2 (n ¼ 380) (83.4–91.8)
Quartile 3 (n ¼ 381) (91.9–101.2)
OR (95% CI)
OR (95% CI)
Quartile 4 (n ¼ 380) (101.2–138.6)
OR (95% CI)
GFR as continuous independent variable P for linear trend
OR per 10 ml/min per 1.73 m2 increment (95% CI)
(0.81–1.60) (0.84–1.70) (0.84–1.70) (0.83–1.69)
1.11 1.12 1.12 1.11
(0.79–1.58) (0.78–1.60) (0.78–1.60) (0.77–1.59)
1.49 1.45 1.45 1.43
(1.05–2.09)a (1.02–2.07)a (1.02–2.07)a (1.00–2.04)a
0.01 0.02 0.02 0.02
1.14 1.13 1.13 1.12
(1.26–1.04) (1.25–1.02) (1.25–1.02) (1.24–1.02)
Total plaque area greater than the median of non-zero values Model 1 1.00 (Reference) 0.67 (0.42–1.05) Model 2 1.00 (Reference) 0.70 (0.44–1.12) Model 3 1.00 (Reference) 0.70 (0.44–1.13) Model 4 1.00 (Reference) 0.71 (0.45–1.14)
0.86 0.86 0.86 0.88
(0.55–1.35) (0.54–1.35) (0.55–1.37) (0.55–1.39)
1.55 (1.03–2.34)a 1.52 (1.00–2.32)a 1.52 (1.00–2.32) 1.56 (1.02–2.39)a
0.01 0.03 0.03 0.02
1.16 1.14 1.14 1.15
(1.31–1.03) (1.29–1.01) (1.29–1.01) (1.30–1.02)
(1.07–2.27)a (1.05–2.28)a (1.05–2.28)a (1.10–2.38)a
0.05 0.05 0.05 0.03
1.11 1.11 1.12 1.13
(1.00–1.24) (1.00–1.24) (1.00–1.24) (1.01–1.26)
1.14 1.20 1.20 1.19
Left ventricular hypertrophy by either the Sokolow voltage, Cornell product, or MESA voltage criterion Model 1 1.00 (Reference) 1.05 (0.71–1.56) 1.38 (0.95–2.01) Model 2 1.00 (Reference) 1.00 (0.67–1.50) 1.40 (0.95–2.05) Model 3 1.00 (Reference) 1.01 (0.68–1.51) 1.41 (0.96–2.07) Model 4 1.00 (Reference) 1.03 (0.69–1.54) 1.44 (0.98–2.13)
1.56 1.55 1.55 1.62
Abbreviations: CI, confidence interval; GFR, glomerular filtration rate; OR, odds ratio; MESA, Multi-Ethnic Study of Atherosclerosis; RENIS-T6, Renal Iohexol Clearance Survey in Tromsø 6. Model 1 was adjusted for age, sex, body weight, height, body mass index, and the use of lipid-lowering and nonsteroidal anti-inflammatory drugs, angiotensin receptor blockers, angiotensin-converting enzyme inhibitors, b-blockers, calcium blockers, diuretics, and other hypertensive antiinflammatory drugs. Model 2 was adjusted as model 1 and for cardiovascular risk factors (first-degree relative with myocardial infarction before 60 years of age, physical exercise, number of cigarettes currently smoked, ambulatory daytime mean arterial pressure, ambulatory daytime pulse pressure, low-density lipoprotein (LDL) cholesterol, high-density lipoprotein (HDL) cholesterol, fasting triglycerides, and serum high-sensitivity C-reactive protein). Model 3 was adjusted as model 2 and for urinary albumin–creatinine ratio. Model 4 was adjusted as model 3 and for fasting serum glucose. a Po0.05 for difference from quartile 1.
survey (October 2007 to December 2008) included a 40% random sample of individuals aged 50 to 59 years drawn from the population registry and all individuals aged 60 to 62 years (5464 total subjects). A total of 3564 subjects aged 50–62 years completed the survey. We excluded 739 subjects with prevalent myocardial infarction, angina pectoris, stroke, diabetes mellitus, or renal disease. The remaining 2825 eligible subjects were invited to the RENIS-T6. Of the 2107 who responded positively, 12 were excluded because of an allergy to contrast media, iodine, or latex; 65 because of other reasons; and 48 did not present for their appointments. A total of 1982 subjects remained for potential inclusion, and 1632 were investigated according to a predefined target. Five participants were excluded because of technical failures in their GFR measurements, leaving 1627 included participants in the RENIS-T6 cohort (Figure 1). The cohort has previously been shown to be representative of all 2825 eligible subjects.20 For the present analyses, we excluded individuals with chronic kidney disease (GFR o60 ml/min per 1.73 m2; or urinary ACR 41.92 mg/mmol for men and 2.83 mg/mmol for women)34 or diabetes (fasting serum glucose X7.0 mmol/l; or hemoglobinA1C X6.5%).35 A questionnaire about CVD risk factors was administered. Smoking status was divided into current smokers or nonsmokers. The number of cigarettes currently smoked was registered. A family history of early myocardial infarction was defined as a first-degree relative with myocardial infarction before the age of 60 years. Physical exercise was defined as leisure-time activity leading to perspiration or breathlessness. This study complied with the Declaration of Helsinki Principles and was approved by the Norwegian Data Inspectorate 150
and the Regional Committee for Medical and Health Research Ethics of North Norway. All subjects provided written consent. Carotid ultrasonography High-resolution B-mode ultrasonography was performed with a duplex scanner (GE Vivid 7, GE Healthcare, Pittsburgh, PA) with a 12 MHz linear array transducer. The sonographers underwent a 2-month training program before study start, and standard operational procedures were used to minimize errors. Readings were centralized and standardized. The far and near wall of the right common carotid artery, bifurcation, and internal carotid artery (6 locations) were scanned for plaques. A plaque was defined as a localized protrusion with thickening of the vessel wall of 450% compared with the adjacent IMT. Each plaque was outlined on still images with calculation of TPA as the sum of all plaque areas. Automated measurement of the IMT was performed in 10 mm segments on three separate images from the far and near wall of the common carotid artery and the far wall of the bifurcation. The mean IMT from the three images were calculated for each location, and the average of the means in all three locations (mean IMT) was used in the analyses. If present in the predefined location of interest, plaques were included in the IMT measurements. The intra- and interobserver reproducibility were acceptable. Further details have been published previously.36 Electrocardiographic assessment of LVH A standard 12-lead ECG was performed with a Cardiovit AT-104 PC Electrocardiograph by trained medical students and nurses (Schiller AG, Baar, Switzerland). The amplitudes and durations of the different deflections were output automatically by the system. As Kidney International (2014) 86, 146–153
clinical investigation
BO Eriksen et al.: Subclinical cardiovascular disease and high GFR
recommended in current guidelines, more than one diagnostic criterion for LVH was used.37 We applied the Sokolow voltage and the Cornell product,38 as well as the MESA voltage criterion, which has recently been shown to perform well.39 LVH was judged to be present when detected by any of these three criteria. The Sokolow voltage was calculated as the sum of SV1 and the highest of either RV5 or RV6.40 The Cornell product was calculated as the product of the QRS duration and the sum of RaVL and SV3 plus 0.8 for women.38 The MESA voltage was calculated as the sum of SV1, SV2, and RV5.39 The thresholds indicating LVH were X3.5 mV, X244 mV ms, and X4.2 mV, respectively. Iohexol clearance GFR was measured as the single-sample plasma clearance of iohexol. This method has been validated against gold standard methods. Five milliliters of iohexol (Omnipaque, 300 mg I/ml, Amersham Health, London, UK) was injected intravenously. The exact time was measured to sampling for measurement of iohexol concentration by high-performance liquid chromatography. Details have been published previously.20 Blood pressure Ambulatory BP was measured with the Spacelab 90207 (Spacelab, Redmond, WA). The daytime mean systolic and diastolic ambulatory BPs were calculated as the weighted mean of the measurements from 10.00 to 20.00. Mean pulse pressure was defined as the mean systolic minus the mean diastolic ambulatory BP. Mean arterial pressure was defined as the mean diastolic ambulatory BP plus one-third of the pulse pressure. Individuals with conventional systolic BP 4140 mm Hg, conventional diastolic BP 490 mm Hg, or individuals using antihypertensive medication were categorized as having hypertension. The methods have been described previously.11 Other measurements Fasting serum glucose, triglycerides, and low-density lipoprotein and high-density lipoprotein cholesterol were measured on the Modular model P800 (Roche Diagnostics, Indianapolis, IN). Hemoglobin A1C was measured with a liquid chromatographic method (Variant II instrument, Bio-Rad Laboratories, Hercules, CA). Three consecutive samples of first-void spot urine were analyzed for albumin and creatinine as described previously.41 The ACR was calculated for each specimen, and the median ACR was used in the analyses. Statistical analysis Continuous variables are presented as the median (interquartile range). Differences across GFR quartiles were tested with analysis of variance, median, ordinal logistic, logistic, or Poisson regression as appropriate. Tests for unadjusted linear and quadratic trends were performed using linear and squared terms for GFR in the regressions. Serum triglycerides and mean IMT were logtransformed, and the TPA was square root-transformed to reduce skewness. Adjusted models of mean IMT, TPA, and LVH were analyzed separately with multiple logistic regressions using GFR as the independent variable. For the mean IMT, the dependent dichotomous variable was defined as the upper tertile. Because the majority of the subjects did not have atherosclerotic plaques, the dependent dichotomous variable of TPA was defined as TPA greater Kidney International (2014) 86, 146–153
than the median of non-zero values. The dichotomous LVH variable was defined as LVH detected by the Sokolow voltage, Cornell product, or MESA voltage criteria. The three criteria were also analyzed as the dependent continuous variables in multiple linear regression analyses. Four subjects with left bundle branch block were excluded from the LVH analyses.37 For each of the dependent variables above, separate multiple logistic regression analyses with GFR as an independent variable both divided into quartiles and as a continuous variable were performed. We adjusted for age, gender, body weight, height, body mass index, and the use of lipid-lowering and nonsteroidal anti-inflammatory drugs, angiotensin receptor blockers, angiotensin-converting enzyme inhibitors, b-blockers, calcium blockers, diuretics, and other hypertensives as dichotomous variables (model 1). Model 2 used the same factors as model 1 as well as a family history of early myocardial infarction (yes/no), physical exercise (yes/no), daytime ambulatory mean arterial pressure and pulse pressure, number of currently smoked cigarettes, low-density lipoprotein and high-density lipoprotein cholesterol, the log of fasting triglycerides, and high-sensitivity C-reactive protein. Model 3 used the same factors as model 2 as well as urinary ACR. Model 4 used the same factors as model 3 as well as fasting glucose. Body mass index was calculated as body weight divided by the square of height. Interactions between sex and GFR were tested in all fully adjusted models. Carotid data were missing for 3.6% of the cohort, and ECG data for 16.4%. In accordance with current recommendations, missing data were imputed using multiple chained imputation in STATA/MP 12.1 (StataCorp LP, College Station, TX, www.stata.com; see Supplementary Methods online, including Supplementary Table S1 and Supplementary Figures S1–S6).42 All statistical tests were performed on 50 imputed data sets using mi estimate in STATA, except for the nonlinear analyses described below, which are not supported by mi estimate and were performed for complete cases. Nonlinear associations between the dependent variables and GFR or any of the adjustment variables in the fully adjusted models were analyzed in multivariable second-degree fractional polynomial models with mfp in STATA. GFR was analyzed in the customary way as standardized to 1.73 m2 of body surface area. As this is a disputed issue, all the analyses above were repeated with GFR standardized to 40 l of estimated total body water (GFR40), as suggested by us in a previous publication.20 We also repeated the analyses with absolute (unadjusted) GFR measured in ml/min. Statistical significance was defined as Po0.05, and all statistical tests were two sided. DISCLOSURE
All the authors declared no competing interests. ACKNOWLEDGMENTS
We are deeply grateful for the valuable advice from the late Professor Egil Arnesen, former principal investigator of the Tromsø Study, who was active with this study until his death in December 2009. We thank Bjørg Skog Høgset, Saskia van Heusden, and the staff at the Clinical Research Unit (University Hospital of North Norway) for assistance in planning the study, performing the procedures, and collecting data according to the GCP standard. We also thank Harald Strand and the staff at the Department of Medical Biochemistry (University Hospital of North Norway) for HPLC analyses of iohexol. We also want to thank all the participants in the RENIS-T6 cohort for making this study possible. 151
clinical investigation
SUPPLEMENTARY MATERIAL Table S1. Missing data. Table S2. Multiple logistic regression analyses of electrocardiographic criteria of left ventricular hypertrophy by quartiles of GFR, and by GFR as a continuous variable. Table S3. Multiple linear regression analyses of electrocardiographic criteria of left ventricular hypertrophy by quartiles of GFR, and by GFR as a continuous variable. Table S4. Study population characteristics according to tertile of mean carotid intima-media thickness. Table S5. Study population characteristics according to category of carotid total plaque area. Table S6. Study population characteristics according to left ventricular hypertrophy. Figure S1. Comparison of the cumulative distribution functions of observed and imputed Sokolow voltage for every tenth imputation. Figure S2. Comparison of the cumulative distribution functions of observed and imputed Cornell product for every tenth imputation. Figure S3. Comparison of the cumulative distribution functions of observed and imputed MESA voltage for every tenth imputation. Figure S4. Comparison of the cumulative distribution functions of observed and imputed mean intima-media thickness for every tenth imputation. Figure S5. Comparison of the cumulative distribution functions of observed and imputed total plaque area for every tenth imputation. Figure S6. Comparison of the histograms of observed and imputed number of carotid plaques for every tenth imputation. Supplementary material is linked to the online version of the paper at http://www.nature.com/ki
BO Eriksen et al.: Subclinical cardiovascular disease and high GFR
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23. 24.
REFERENCES Chronic Kidney Disease Prognosis Consortium. Association of estimated glomerular filtration rate and albuminuria with all-cause and cardiovascular mortality in general population cohorts: a collaborative meta-analysis. Lancet 2010; 375: 2073–2081. 2. Nitsch D, Grams M, Sang Y et al. Associations of estimated glomerular filtration rate and albuminuria with mortality and renal failure by sex: a meta-analysis. BMJ 2013; 346: f324. 3. Di Angelantonio E, Chowdhury R, Sarwar N et al. Chronic kidney disease and risk of major cardiovascular disease and non-vascular mortality: prospective population based cohort study. BMJ 2010; 341: c4986. 4. Inker LA, Schmid CH, Tighiouart H et al. Estimating glomerular filtration rate from serum creatinine and cystatin C. N Engl J Med 2012; 367: 20–29. 5. Mathisen UD, Melsom T, Ingebretsen OC et al. Estimated GFR is associated with cardiovascular risk factors independently of measured GFR. J Am Soc Nephrol 2011; 22: 927–937. 6. Knight EL, Verhave JC, Spiegelman D et al. Factors influencing serum cystatin C levels other than renal function and the impact on renal function measurement. Kidney Int 2004; 65: 1416–1421. 7. Rule AD, Bailey KR, Lieske JC et al. Estimating the glomerular filtration rate from serum creatinine is better than from cystatin C for evaluating risk factors associated with chronic kidney disease. Kidney Int 2013; 83: 1169–1176. 8. Den Ruijter HM, Peters SE, Anderson TJ et al. Common carotid intimamedia thickness measurements in cardiovascular risk prediction: a metaanalysis. JAMA 2012; 308: 796–803. 9. Chou R, Arora B, Dana T et al. Screening asymptomatic adults with resting or exercise electrocardiography: a review of the evidence for the U.S. Preventive Services Task Force. Ann Intern Med 2011; 155: 375–385. 10. Wyman RA, Mays ME, McBride PE et al. Ultrasound-detected carotid plaque as a predictor of cardiovascular events. Vasc Med 2006; 11: 123–130. 11. Mathisen UD, Melsom T, Ingebretsen OC et al. Ambulatory blood pressure is associated with measured glomerular filtration rate in the general middle-aged population. J Hypertens 2012; 30: 497–504. 12. Melsom T, Mathisen UD, Ingebretsen OC et al. Impaired fasting glucose is associated with renal hyperfiltration in the general population. Diabetes Care 2011; 34: 1546–1551.
25.
1.
152
26.
27.
28. 29. 30.
31.
32.
33.
34.
35. 36.
37.
Bui AL, Katz R, Kestenbaum B et al. Cystatin C and carotid intima-media thickness in asymptomatic adults: the Multi-Ethnic Study of Atherosclerosis (MESA). Am J Kidney Dis 2009; 53: 389–398. Bobbert T, Mai K, Fischer-Rosinsky A et al. Relation between physiological variation of renal function and carotid intima media thickness in nondiabetic individuals. J Atheroscler Thromb 2010; 17: 242–248. Desbien AM, Chonchol M, Gnahn H et al. Kidney function and progression of carotid intima-media thickness in a community study. Am J Kidney Dis 2008; 51: 584–593. Kastarinen H, Ukkola O, Kesaniemi YA. Glomerular filtration rate is related to carotid intima-media thickness in middle-aged adults. Nephrol Dial Transplant 2009; 24: 2767–2772. Choi SW, Kim HY, Lee YH et al. eGFR is associated with subclinical atherosclerosis independent of albuminuria: the Dong-gu Study. Atherosclerosis 2010; 212: 661–667. Jiang SM, Sun XF, Gu HX et al. Effects of decline in renal function with age on the outcome of asymptomatic carotid plaque in healthy adults: a 5-year follow-up study. Chin Med J 2012; 125: 2649–2657. Moran A, Katz R, Jenny NS et al. Left ventricular hypertrophy in mild and moderate reduction in kidney function determined using cardiac magnetic resonance imaging and cystatin C: the Multi-Ethnic Study of Atherosclerosis (MESA). Am J Kidney Dis 2008; 52: 839–848. Eriksen BO, Melsom T, Mathisen UD et al. GFR normalized to total body water allows comparisons across genders and body sizes. J Am Soc Nephrol 2011; 22: 1517–1525. Delanaye P, Radermecker RP, Rorive M et al. Indexing glomerular filtration rate for body surface area in obese patients is misleading: concept and example. Nephrol Dial Transplant 2005; 20: 2024–2028. Okada R, Yasuda Y, Tsushita K et al. Glomerular hyperfiltration in prediabetes and prehypertension. Nephrol Dial Transplant 2011; 27: 1821–1825. Tomaszewski M, Charchar FJ, Maric C et al. Glomerular hyperfiltration: a new marker of metabolic risk. Kidney Int 2007; 71: 816–821. Maeda I, Hayashi T, Sato KK et al. Cigarette smoking and the association with glomerular hyperfiltration and proteinuria in healthy middle-aged men. Clin J Am Soc Nephrol 2011; 6: 2462–2469. Pinto-Sietsma SJ, Mulder J, Janssen WM et al. Smoking is related to albuminuria and abnormal renal function in nondiabetic persons. Ann Intern Med 2000; 133: 585–591. Verhave JC, Hillege HL, Burgerhof JG et al. The association between atherosclerotic risk factors and renal function in the general population. Kidney Int 2005; 67: 1967–1973. Licht CM, de Geus EJ, Penninx BW. Dysregulation of the autonomic nervous system predicts the development of the metabolic syndrome. J Clin Endocrinol Metab 2013; 98: 2484–2493. Brenner BM, Lawler EV, Mackenzie HS. The hyperfiltration theory: a paradigm shift in nephrology. Kidney Int 1996; 49: 1774–1777. Ruggenenti P, Remuzzi G. Time to abandon microalbuminuria? Kidney Int 2006; 70: 1214–1222. Ford ES, Zhao GX, Li CY. Pre-diabetes and the risk for cardiovascular disease: a systematic review of the evidence. J Am Coll Cardiol 2010; 55: 1310–1317. Hallan SI, Ritz E, Lydersen S et al. Combining GFR and albuminuria to classify CKD improves prediction of ESRD. J Am Soc Nephrol 2009; 20: 1069–1077. Gaspari F, Guerini E, Perico N et al. Glomerular filtration rate determined from a single plasma sample after intravenous iohexol injection: is it reliable? J Am Soc Nephrol 1996; 7: 2689–2693. Bird NJ, Peters C, Michell AR et al. Comparison of GFR measurements assessed from single versus multiple samples. Am J Kidney Dis 2009; 54: 278–288. K/DOQI clinical practice guidelines for chronic kidney disease work group. Part 4. Definition and classification of stages of chronic kidney disease. Am J Kidney Dis 2002; 39: S46–S75. American Diabetes Association. Diagnosis and classification of diabetes mellitus. Diabetes Care 2011; 34: S62–S69. Herder M, Johnsen SH, Arntzen KA et al. Risk factors for progression of carotid intima-media thickness and total plaque area: a 13-year follow-up study: the Tromso Study. Stroke 2012; 43: 1818–1823. Hancock EW, Deal BJ, Mirvis DM et al. AHA/ACCF/HRS recommendations for the standardization and interpretation of the electrocardiogram: part V: electrocardiogram changes associated with cardiac chamber hypertrophy: a scientific statement from the American Heart Association Electrocardiography and Arrhythmias Committee, Council on Clinical Cardiology; the American College of Cardiology Foundation; and the Heart Rhythm Society Endorsed by the International Kidney International (2014) 86, 146–153
clinical investigation
BO Eriksen et al.: Subclinical cardiovascular disease and high GFR
38.
39.
Society for Computerized Electrocardiology. J Am Coll Cardiol 2009; 53: 992–1002. Devereux RB, Bella J, Boman K et al. Echocardiographic left ventricular geometry in hypertensive patients with electrocardiographic left ventricular hypertrophy: the LIFE study. Blood Pressure 2001; 10: 74–82. Jain A, Tandri H, Dalal D et al. Diagnostic and prognostic utility of electrocardiography for left ventricular hypertrophy defined by magnetic resonance imaging in relationship to ethnicity: the Multi-Ethnic Study of Atherosclerosis (MESA). Am Heart J 2010; 159: 652–658.
Kidney International (2014) 86, 146–153
40.
41.
42.
Sokolow M, Lyon TP. The ventricular complex in left ventricular hypertrophy as obtained by unipolar precordial and limb leads. Am Heart J 1949; 37: 161–186. Solbu MD, Kronborg J, Eriksen BO et al. Cardiovascular risk-factors predict progression of urinary albumin-excretion in a general, non-diabetic population: a gender-specific follow-up study. Atherosclerosis 2008; 201: 398–406. Little RJ, D’Agostino R, Cohen ML et al. The prevention and treatment of missing data in clinical trials. N Engl J Med 2012; 367: 1355–1360.
153
