Cystatin C and risk of vascular and nonvascular mortality: a prospective cohort study of older men

Authors


Dr Jonathan Emberson, Clinical Trial Service Unit and Epidemiological Studies Unit, Richard Doll Building, University of Oxford, Old Road Campus, Roosevelt Drive, Oxford OX3 7LF, UK.
(fax: +44-1865-743985; e-mail: jonathan.emberson@ctsu.ox.ac.uk).

Abstract

Abstract.  Emberson JR, Haynes R, Dasgupta T, Mafham M, Landray MJ, Baigent C, Clarke R (University of Oxford, Oxford, UK). Cystatin C and risk of vascular and nonvascular mortality: a prospective cohort study of older men. J Intern Med 2010; 268: 145–154.

Objective.  To assess the relevance of cystatin C, as a marker of mild-to-moderate renal impairment, for vascular and nonvascular mortality in older people.

Design.  Prospective cohort study.

Setting.  Re-survey in 1997 to 1998 of survivors in the 1970 Whitehall study of London civil servants.

Subjects.  Five thousand three hundred and seventy-one men (mean age at resurvey: 77 years) who took part in the resurvey and had plasma cystatin C concentration measured.

Main outcome measures.  Cause-specific mortality over subsequent 11 years (1997 to 2008).

Methods.  Cox regression was used to estimate the associations of cystatin C with vascular and nonvascular mortality, before and after adjustment for prior disease and other risk factors (including lifetime blood pressure).

Results.  During an 11.0-year follow-up period, there were 1171 deaths from vascular causes [26 per 1000 per year (py)] and 1615 deaths from nonvascular causes (36 per 1000 py). Compared with men with cystatin C in the bottom fifth of the distribution, men in the top 10th had about two-fold higher mortality rates from vascular and nonvascular mortality (fully adjusted P both <0.001) even after adjustment for prior disease and all measured confounders, including lifetime blood pressure. The fully adjusted relative risks per 50% higher cystatin C concentrations were 1.66 [95% CI 1.48 to 1.85] for vascular mortality, 1.92 [95% CI 1.66 to 2.22] for ischaemic heart disease mortality and 1.46 [95% CI 1.31 to 1.61] for nonvascular mortality.

Conclusions.  In older men, plasma concentration of cystatin C, probably as a marker of mild renal disease, is a strong independent predictor of both vascular and nonvascular mortality.

Introduction

Death rates amongst people with chronic kidney disease requiring dialysis are about 10 to 100 times greater than those in the general population, with much of this excess mortality being from vascular causes [1]. Chronic kidney disease (CKD) [defined as an estimated glomerular filtration rate (eGFR) of <60 mL per min per 1.73 m2 body surface area when measured on two occasions after a 3 month interval] [2], is also associated with a substantial excess mortality from vascular and from nonvascular causes [3–6]. Several population-based studies of healthy people, using mainly creatinine-based methods to estimate GFR, have reported that people with CKD have 30% to 80% higher risks of all-cause mortality [3–6]. Whilst the absolute risk of death increases exponentially with decreasing renal function, there is considerable uncertainty about the shape and strength of any associations of mild-to-moderate renal impairment with vascular and nonvascular mortality. In particular, it is unclear if the risks of vascular disease increase only when renal function drops below a certain threshold. Some of the uncertainties about the shape and strength of the associations of renal function with vascular diseases relate to an insufficient number of events examined in previous studies, variable inclusion of people with prior vascular disease at enrolment, incomplete adjustment for other risk factors and the insensitivity of creatinine as a marker of renal function in people with normal or mildly impaired renal function [7–10].

Since renal function declines with age and the absolute risks of death and disease increase with age, there is a need to assess the relevance of mild-to-moderate reductions in renal function for cause-specific mortality in older people. Such an assessment should take account of known cardiovascular risk factors (including lifetime blood pressure) and avoid reliance on creatinine-based estimates of GFR. Cystatin C, a nonglycosylated basic protein produced at a constant rate by all nucleated cells, has been proposed as a more sensitive marker of renal function than creatinine and particularly in the setting of mild renal impairment [10–12]. Cystatin C is filtered by the renal glomerulus and metabolized by the proximal tubule. In contrast with creatinine, blood cystatin C concentrations appear to be much less affected by age, sex or muscle mass. The aims of this study were to assess the shape and strength of the associations of cystatin C with vascular and nonvascular mortality, after taking account of prior disease and other risk factors, in a prospective cohort study of 5371 older men followed for an average of 11 years.

Materials and methods

Study population

Between 1967 and 1970, 19 019 male civil servants aged 40–69 years working in London were recruited into the Whitehall study [13, 14]. All 8448 surviving participants were sought for re-survey in 1997–1998 [15, 16]. A postal questionnaire asked about medical history (diagnoses of vascular diseases and cancer), medications taken in the last month, smoking and other lifestyle characteristics and last known civil service employment grade [16]. The 7044 (83%) men who responded to the re-survey questionnaire were sent a blood collection kit and asked to attend their local family doctors to have a blood sample collected and blood pressure measured. Urine samples were not requested. Of the 5434 men who returned a nonfasting blood sample (77% of respondents, 64% of those alive), cystatin C was measured and subsequent mortality follow-up (see below) was ascertained for 5371 (98.8%). The re-survey was approved by the ethics committees of the participating institutions.

Laboratory methods

Blood was collected into a 10 mL vacutainer containing potassium EDTA with 0.34 mmol L−1 aprotinin. These nonfasting whole blood samples were mailed in sealed transport tubes at room temperature to the CTSU Wolfson laboratories in Oxford. The mean time in the post was 1.3 days (range 0–7 days), with 78% arriving within 24 h and 96% arriving within 48 h of blood collection. On arrival in the laboratory, the blood was centrifuged and the plasma was aliquoted for storage at −40 °C. Cystatin C was measured using Dade Behring reagents, on an automated Dade Behring BN II nephelometer (Siemens Health Care Diagnostics, Camberley, UK). Total cholesterol, directly measured low-density lipoprotein cholesterol [LDL-C], high-density lipoprotein cholesterol [HDL-C], apolipoprotein A1 [Apo A1] and apolipoprotein B [Apo B] as well as markers of inflammation [C-reactive protein (CRP), fibrinogen and albumin] were measured using standard methods [17, 18] (but creatinine was not measured). The inter-assay coefficient of variation for cystatin C, based on repeat analysis of laboratory control material, was 5%. Previous studies had indicated that minor changes in some blood assays can arise due to delayed separation of mailed blood samples; so, wherever possible (>99% of samples), cystatin C values were adjusted for duration of time spent in the post before separation of plasma [19]; and also for date of assay to correct for any assay drift. Estimated glomerular filtration rate (eGFR, in mL min−1 1.73 m−2) was calculated from cystatin C concentration using the formula inline image [20].

Mortality follow-up

Participants were flagged for mortality at the Office for National Statistics (England), which provided the date and cause [including International Classification of Disease (ICD) codes] of all deaths occurring until the end of August 2008. The mean follow-up period amongst survivors was 11.0 years. Cause-specific mortality was coded using ICD-9 up to August 2002 and ICD-10 subsequently. Vascular deaths (heart disease, stroke and other vascular disease) were so defined if coded as ICD-9 codes 390-459, 798 or ICD-10 codes I00-I99, R96, with all other causes of death being defined as nonvascular. Deaths from ischaemic heart disease (IHD; ICD-9 codes 410-414; ICD-10 codes I20-I25), stroke (ICD-9 codes 430-438; ICD-10 codes I60-I69), cancer (ICD-9 codes 140-208; ICD-10 codes C0-C97) and respiratory disease (ICD-9 codes 460-519; ICD-10 codes J00-J99) were also analysed separately.

Statistical methods

In order to assess the shape of the association of cystatin C with mortality, men were classified into one of five equally sized groups (I–V; each comprising about 1000 men) according to their baseline cystatin C concentration, with the top group being further subdivided in two (Va and Vb; corresponding to the top two tenths of the distribution). The mean and prevalence of other baseline characteristics in each of these groups were age-adjusted (using standard regression techniques) and tests for trend across the groups were performed by including log cystatin C in the appropriate regression model. The association of cystatin C with mortality was assessed using Cox proportional hazards regression. Relative risks (RRs: approximated by the hazard ratio in the Cox models) were estimated for each group relative to group I and are shown as ‘floating absolute risks’ (which does not alter their values but merely ascribes an approximate 95% confidence interval to the RR in every group) [21]. In order to assess the strength of the association of cystatin C with mortality, the average risk observed across the range of cystatin C levels studied was estimated by including measured log cystatin C in the Cox model as a linear term, and estimating the RR per 50% higher cystatin C concentration (approximately a 2 SD difference on the log scale, corresponding to the difference in the mean concentration between the top and bottom thirds). Tests for deviation from this ‘log-linear’ model were performed by assessing the statistical significance of an additional quadratic term for log cystatin C.

The analyses were repeated after adjustment for age, age plus prior chronic disease (myocardial infarction or angina, stroke, diabetes, cancer), and age plus prior chronic disease and vascular risk factors other than blood pressure [current or ex smoking, current alcohol drinking, employment grade, lipids and apolipoproteins (LDL-C, HDL-C, Apo A1, Apo B), and markers of inflammation (C-reactive protein, albumin, fibrinogen)]. Additional analyses also included further adjustment for ‘lifetime blood pressure’, defined as a recall of a diagnosis of hypertension, use of blood pressure lowering drugs, and measured systolic and diastolic blood pressure not only at re-survey but also at entry to the Whitehall study in 1967 to 1970. The difference in minus twice the log likelihood statistic between nested models that include and exclude log cystatin C provides a statistical test of the predictive value of log cystatin C (over and above other factors in that model) and, by comparison between different models, can also be used to provide an indication of the extent to which any apparent risk associations (e.g. in unadjusted models) are explained by other factors. To assess the extent that any associations of cystatin C with mortality could be due to reverse causality, the analyses carried out in all men were repeated separately in men with and without a prior history of cardiovascular disease (CVD), defined as a recall of a diagnosis of angina, myocardial infarction or stroke at re-survey. (Four equally sized groups were used when displaying floating absolute risks separately in men with and without prior CVD for statistical stability when assessing sub-group analyses.) In addition, as a further assessment of the effects of reverse causality, risk associations per 50% higher cystatin C concentration were estimated separately during the first 6 years of follow-up (during which about half of all deaths occurred) and in later years, and a test of difference between the two estimated hazard ratios was performed.

Results

Characteristics of study population by baseline cystatin C and prior CVD

Amongst the 5371 men with data on cystatin C and mortality follow-up, 1341 (25.0%) had prior CVD (myocardial infarction, angina or stroke). Men in the top fifth of the distribution were on average about 5 years older than men in the bottom fifth (Table 1). At any given age, men with higher levels of cystatin C were more likely to have a prior diagnosis of IHD, stroke, diabetes and cancer than men with lower levels, and were much more likely to have been diagnosed with hypertension or to be taking blood pressure-lowering medication. Thus, although measured blood pressure at each participant’s original entry in the Whitehall study (in 1967–70, some 30 years earlier) was strongly related to cystatin C, measured blood pressure at re-survey varied little with cystatin C. Men with higher cystatin C concentrations also had unfavourable levels of LDL : HDL cholesterol, apolipoprotein B : apoliopoprotein A1, C-reactive protein, fibrinogen and albumin.

Table 1. Study characteristics, overall and by prior vascular disease and baseline cystatin C concentration
Baseline characteristicAll menNo prior CVDPrior CVDFifth of baseline cystatin CaPb
IIIIIIIVVaVb
  1. Mean (SD), median (IQR) or n (%) shown.

  2. IHD, ischaemic heart disease (recall of diagnosis of myocardial infarction or angina).

  3. aWith the exception of age, cystatin C and eGFR, estimates are adjusted for age differences across cystatin C groups. bTest of linear trend between log(cystatin C) concentration and baseline characteristic after adjustment for age. cManufacturers reference range for cystatin C: 0.53–0.95 mg L−1. Laboratory 95% reference ranges, based on an external sample of men and women aged 30–79 without a history of cardiovascular disease (controls from the ISIS case–control study) [19]: LDL-C (1.46–4.36); HDL-C (0.49–2.08); apolipoprotein A1 (0.92–1.69); apolipoprotein B (0.60–1.53); C-reactive protein (0.17–16.2); albumin (34.1–46.5); fibrinogen (6.38–13.79).

Number of men5371403013411074107410741076536537
Age, years76.9 (4.9)76.6 (4.8)77.6 (4.9)74.7 (3.8)75.8 (4.5)76.9 (4.8)77.4 (4.8)78.9 (5.0)80.1 (5.0)
Cystatin C, mg L−1 c1.04 (0.93–1.20)1.02 (0.92–1.17)1.11 (0.97–1.31)0.83 (0.05)0.95 (0.03)1.04 (0.03)1.16 (0.05)1.32 (0.05)1.77 (0.50)
eGFR, mL min−1 1.73 m−272.2 (15.3)73.7 (14.7)67.6 (16.3)92.6 (6.7)80.3 (2.4)73.0 (2.0)65.0 (2.8)56.7 (2.1)43.3 (8.3)
Medical history, %
 IHD20.00.079.915.013.417.521.027.032.9<0.001
 Stroke7.10.028.64.34.66.27.19.412.8<0.001
 Diabetes5.94.69.85.65.54.55.56.79.20.001
 Hypertension31.727.444.520.624.929.332.344.453.8<0.001
 Cancer (not skin)7.87.78.26.37.77.77.29.49.70.02
Medication, %
 Antiplatelets33.022.165.630.530.731.434.837.835.80.001
 Blood pressure lowering drugs32.024.355.018.922.026.534.446.163.3<0.001
 Statins2.30.86.92.22.01.72.82.04.00.15
Lifestyle, %
 Current tobacco smoker12.813.99.58.810.413.115.016.814.2<0.001
 Current alcohol drinker77.978.874.984.579.178.777.073.669.9<0.001
Blood pressure, mm Hg
 Systolic blood pressure144.9 (20.2)145.4 (20.0)143.2 (20.5)144.5 (20.5)145.4 (20.2)144.4 (20.1)146.2 (20.1)144.8 (20.3)142.6 (20.6)0.10
 Diastolic blood pressure80.2 (10.8)80.7 (10.7)78.5 (10.8)79.8 (11.0)80.6 (10.8)80.2 (10.7)80.6 (10.8)80.5 (10.8)78.9 (11.0)0.07
Blood pressure measured in 1967–1970 (∼30 years earlier), mm Hg
 Systolic blood pressure130.8 (17.6)129.7 (17.2)134.1 (18.3)129.1 (17.8)129.6 (17.5)130.4 (17.4)130.4 (17.5)132.9 (17.6)135.9 (17.9)<0.001
 Diastolic blood pressure80.3 (12.1)79.6 (11.8)82.4 (12.7)78.7 (12.3)79.6 (12.1)80.1 (12.0)80.3 (12.0)82.2 (12.1)83.3 (12.3)<0.001
Laboratory measurementsc
 LDL-C, mmol L−13.37 (0.79)3.38 (0.78)3.34 (0.79)3.33 (0.80)3.41 (0.78)3.41 (0.78)3.37 (0.78)3.33 (0.79)3.33 (0.80)0.62
 HDL-C, mmol L−11.09 (0.38)1.12 (0.38)1.02 (0.36)1.24 (0.37)1.16 (0.37)1.09 (0.36)1.04 (0.36)0.97 (0.37)0.90 (0.37)<0.001
 LDL-C/HDL-C3.5 (1.9)3.4 (1.6)3.8 (2.5)3.0 (1.9)3.2 (1.8)3.6 (1.8)3.7 (1.8)4.1 (1.8)4.3 (1.9)<0.001
 Apolipoprotein A1, g L−10.95 (0.14)0.96 (0.15)0.92 (0.14)1.00 (0.14)0.97 (0.14)0.95 (0.14)0.94 (0.14)0.91 (0.14)0.90 (0.15)<0.001
 Apolipoprotein B, g L−10.87 (0.23)0.86 (0.23)0.88 (0.23)0.82 (0.23)0.86 (0.23)0.87 (0.23)0.88 (0.23)0.89 (0.23)0.91 (0.23)<0.001
 Apo B/Apo A10.93 (0.30)0.92 (0.29)0.97 (0.30)0.85 (0.30)0.90 (0.29)0.94 (0.29)0.96 (0.29)1.00 (0.29)1.03 (0.30)<0.001
 C-reactive protein, mg L−13.7 (7.7)3.5 (7.8)4.3 (7.3)2.3 (7.7)2.6 (7.6)3.3 (7.5)3.9 (7.5)5.6 (7.6)7.3 (7.7)<0.001
 Albumin, g L−139.7 (2.9)39.8 (2.8)39.4 (3.0)40.2 (2.8)40.0 (2.8)39.8 (2.8)39.6 (2.8)39.4 (2.8)38.2 (2.9)<0.001
 Fibrinogen, μmol L−110.4 (2.5)10.3 (2.5)10.7 (2.5)9.8 (2.5)10.0 (2.4)10.3 (2.4)10.5 (2.4)10.9 (2.4)12.0 (2.5)<0.001

Cystatin C and vascular mortality

During a mean follow-up amongst survivors of 11.0 years (maximum 11.3 years), there were 1171 deaths from vascular causes [26 per 1000 per year (py)], including 589 IHD deaths (13 per 1000 py) and 315 stroke deaths (7 per 1000 py) (Table 2). Vascular mortality rates were higher at higher plasma concentrations of cystatin C irrespective of prior history of CVD (see Fig. 1 and Table S1), with men in the top 10th of the distribution having, at any given age, 3.4 times the average risk of men in the bottom fifth of the distribution (Fig. 1 and Table S2). Across the range of values studied, each 50% higher cystatin C concentration was, on average and at any given age, associated with a 2.1 fold higher vascular mortality risk (95% CI 1.9–2.4; Fig. 2). This relative risk was reduced to 1.78 (95% CI 1.59–1.98) after adjustment for prior disease and confounding risk factors, and to 1.66 (95% CI 1.48–1.85) after additional adjustment for blood pressure. The associations of log cystatin C with vascular mortality after adjustment for known risk factors were approximately ‘log-linear’ throughout the range of values of cystatin C studied (P-value for quadratic term 0.25). Relative risks for IHD mortality were stronger than those for stroke and possibly also for other vascular causes (fully adjusted RR per 50% higher cystatin C concentration: 1.92 [95% CI 1.66–2.22] for IHD, 1.20 [95% CI 0.95–1.52] for stroke and 1.66 [95% CI 1.30–2.12] for other vascular causes; Fig. 3).

Table 2. Cause-specific mortality: number of deaths (annual death rate per 1000 men), overall and by prior vascular disease
 All menNo prior CVDPrior CVD
  1. IHD, ischaemic heart disease.

Number of men537140301341
Total follow-up, person-years45 27235 5609712
Cause of death
 IHD589 (13.0)300 (8.4)289 (29.8)
 Stroke315 (7.0)203 (5.7)112 (11.5)
 Other vascular267 (5.9)191 (5.4)76 (7.8)
Subtotal: any vascular cause1171 (25.9)694 (19.5)477 (49.1)
 Cancer729 (16.1)545 (15.3)184 (18.9)
 Respiratory427 (9.4)297 (8.4)130 (13.4)
 Other nonvascular459 (10.1)342 (9.6)117 (12.0)
Subtotal: any nonvascular cause1615 (35.7)1184 (33.3)431 (44.4)
Total: any cause2786 (61.5)1878 (52.8)908 (93.5)
Figure 1.

Age-adjusted relevance of measured cystatin C for vascular and nonvascular mortality in old age, overall and separately in men with and without prior cardiovascular disease. CVD, cardiovascular disease (recall of a diagnosis of myocardial infarction, angina or stroke); eGFR, estimated glomerular filtration rate.

Figure 2.

Effect of adjustment for known risk factors on the association between cystatin C and vascular and nonvascular mortality. arecall of a diagnosis of ischaemic heart disease, stroke, cancer or diabetes. bsmoking status (current/ex/never); drinking status (current/non); grade of employment; LDL–C, HDL–C, apo A1, apo B; markers of inflammation (albumin, fibrinogen and C–reactive protein). crecall (in 1997) of a diagnosis of hypertension, treatment (in 1997) for hypertension and measured systolic and diastolic blood pressure in both 1997 and in ∼1970.

Figure 3.

Association between cystatin C and cause-specific mortality after adjustment for measured confounders and for blood pressure. IHD, ischaemic heart disease. Analyses are adjusted for smoking status (current/ex/never), drinking status (current/non), recall of a diagnosis of IHD, stroke, cancer or diabetes, grade of employment, LDL–C, HDL–C, apo A1, apo B, markers of inflammation (albumin, fibrinogen and C-reactive protein), recall (in 1997) of a diagnosis of hypertension, treatment (in 1997) for hypertension and measured systolic and diastolic blood pressure in both 1997 and in ∼1970.

Cystatin C and nonvascular mortality

One thousand six hundred and fifteen men died from nonvascular causes (36 per 1000 py), including 729 from cancer (16 per 1000 py) and 427 from respiratory disease (9 per 1000 py). Nonvascular mortality rates were higher at higher plasma concentrations of cystatin C, with men in the top tenth of the distribution having, at any given age, 2.3-fold greater risks compared with men in the bottom fifth of the distribution (Fig. 1 and Table S2). Across the range of values studied, each 50% higher cystatin C concentration was, on average and at any given age, associated with a 1.7-fold higher risk of nonvascular mortality (95% CI 1.5–1.8) (Fig. 2). This relative risk was reduced to 1.45 (95% CI 1.31–1.60) after adjustment for prior disease and confounding risk factors and was unaffected by further adjustment for blood pressure (RR = 1.46, 95% CI 1.31–1.61). In contrast with vascular mortality, there was some evidence that these fully adjusted risks might not represent a truly log-linear relationship with nonvascular mortality throughout the range of values studied (P-value for quadratic term 0.04). Relative risks for cancer mortality appeared to be weaker than for other nonvascular causes (fully adjusted RRs per 50% higher concentration: 1.21 [95% CI 1.03–1.42] for cancer, 1.56 [95% CI 1.29–1.89] for respiratory and 1.77 [95% CI 1.48–2.13] for other nonvascular causes; Fig. 3).

Sensitivity analyses

To investigate the extent to which the relative risks might be due to reverse causality, the fully adjusted relative risks associated with 50% higher cystatin C concentrations were estimated separately for the first 6 years of follow-up (during which about half of all deaths occurred) and for subsequent years. For both vascular and nonvascular mortality, there were no significant differences between the relative risks estimated in these two periods. For vascular mortality, the fully adjusted relative risk was 1.71 (95% CI 1.50–1.95) during the first 6 years and 1.57 (95% CI 1.32–1.86) for subsequent years (P for difference 0.40), whilst for nonvascular mortality, the fully adjusted relative risks were 1.53 (95% CI 1.36–1.73) and 1.35 (95% CI 1.16–1.56) respectively (P for difference = 0.15). Exclusion of the 1341 men with prior vascular disease from all analyses resulted in some slight attenuation of the adjusted relative risk of vascular mortality (and its components) per 50% higher plasma concentration of cystatin C (RR = 1.53, 95% CI 1.30–1.80) but not for nonvascular mortality (RR = 1.42, 95% CI 1.25–1.61) or its components (see Figures S1 and S2).

Discussion

In the present study, higher plasma cystatin C concentrations were associated with higher vascular and nonvascular mortality rates throughout the range studied and these associations were independent of age, prior disease and known vascular risk factors. Compared with men in the bottom fifth of the cystatin C distribution, otherwise similar men in the top 10th of the distribution had approximately two-fold higher rates of vascular and nonvascular mortality. After adjustment for known risk factors each 50% higher cystatin C concentration was on average associated with about a 70% higher vascular and a 50% higher nonvascular mortality risk.

Validity of findings

Whilst the excess risks associated with high cystatin C in the present report were largely unexplained by established prior vascular disease, they could, in part, be explained by subclinical vascular disease (which is common in the elderly) leading to a reduction in renal function [22, 23]. Alternatively, because the characteristics included in the adjusted models would have been subject to random errors, some residual confounding may well remain. For instance, it is possible that adjustment for measured blood pressure attenuated the relationship between cystatin C and vascular mortality only partly and that full adjustment for ‘usual’ blood pressure would have attenuated the relationship to a greater extent [24] (though adjustment for blood pressure at two points in time in the current paper would have substantially reduced the potential for this to be a major source of bias). Similarly, adjustment for prior diseases may have been incomplete as these were based on self-reported diagnoses (which can be subject to misclassification). On the other hand, cystatin C would also have been measured with some error, and so the risk associations per 50% higher level shown in this paper could equally be underestimates of the true associations with ‘usual’ cystatin C concentrations because of regression dilution bias [24]. The association between higher cystatin C concentrations and higher risk could also be due, in part, to a positive correlation existing between cystatin C and the rate of change in cystatin C [25]. Specifically, the excess risks observed in men with higher baseline cystatin C levels may be due to the fact that men whose renal function was already declining at a clinically important rate would be over-represented in this group [26]. Other unmeasured risk factors (such as proteinuria) might also help explain some of the associations observed for cystatin C.

Comparison with previous studies

Previous studies have also demonstrated the potential relevance of cystatin C for risk prediction purposes in elderly people. In the Cardiovascular Health Study of 4637 elderly men and women followed for an average of 7 years, all-cause mortality rates were more than two-fold higher amongst those in the top fifth of the cystatin C distribution compared with those in the bottom fifth (after adjustment for other characteristics) [27]. Similarly, in the Heart and Soul Study (a study of 990 patients with coronary heart disease), 3-year mortality was more than three-fold higher amongst those in the top quarter of the cystatin C distribution compared with those in the bottom quarter [28], whilst in the Health, Aging and Body Composition Study (a study of 3075 ‘essentially healthy’ people) 6-year mortality was more than two-fold higher amongst those in the top fifth of the cystatin C distribution compared with the bottom fifth [29]. Similar relative risks have also been observed in a recent Japanese prospective study of 1034 people aged 65 or older [30]. Whilst many previous prospective studies have examined relationships between plasma creatinine concentrations and mortality risk, direct comparison with these studies is difficult, not least because different studies tend to use different cut points to define impaired renal function. Whilst we cannot comment on the relative predictive strengths of creatinine and cystatin C in the current study (because creatinine was not measured), in one previous study, cystatin C was found to be a stronger predictor of both nonfatal and fatal cardiovascular disease events than creatinine [27].

The present study confirms the findings of these previous studies of cystatin C and vascular disease, but the much larger number of deaths provides greater precision about the risks of vascular disease associated with extreme elevations of cystatin C and, more importantly, about the associations with mortality at moderately elevated concentrations of cystatin C. Moreover, the present study confirms earlier reports of associations of cystatin C with nonvascular mortality. Indeed, the magnitude of the excess risks for nonvascular mortality associated with high cystatin C concentrations – particularly for respiratory and other non-neoplastic causes of deaths – were greater than expected. Unreported or subclinical nonvascular co-morbidity (e.g. recurrent infections, inflammatory conditions or cancer) and their treatment may adversely affect kidney function as well as increase nonvascular mortality. However, the observed associations with nonvascular mortality remained even after exclusion of those with a history of cardiovascular disease and of the first 6 year of follow-up. There could also be some misclassification of the causes of death, but the magnitude of the effect size and the number of deaths suggests that this is unlikely to be a major influence on the findings. Finally, there may be a causal link between renal impairment and nonvascular mortality (perhaps mediated via immunosuppression) [31], although the association with such a variety of causes of nonvascular death (including cancer, respiratory and all other nonvascular deaths) argues against this.

Use of cystatin C to estimate glomerular filtration rate

Most large scale epidemiological studies estimate GFR from creatinine using either the Modification of Diet in Renal Disease [MDRD] [7, 8] or the Cockcroft-Gault formulas [32]. However, these equations have been found to perform poorly in people with normal, or only mildly impaired, renal function [9, 33]. In the present study, use of cystatin C to estimate GFR yielded average eGFR levels of about 50 mL min−1 1.73 m−2 amongst men in the top fifth of the cystatin C distribution and 90 mL min−1 1.73 m−2 amongst men in the bottom fifth (see Table 1). Whilst factors other than renal function affect cystatin C concentration (as they do creatinine), it has been estimated that about 70% of the variation in cystatin C is due to true differences in measured (i.e. ‘gold-standard’) GFR [34]. This suggests that the associations seen with cystatin C after adjustment for confounders do largely reflect effects of renal function on mortality. However, even if other factors do, even partly, explain these associations, the value of using cystatin C to discriminate mortality risk between older men is undiminished.

Conclusions

The present study demonstrates that cystatin C measured in old age is positively associated not only with the risk of vascular mortality but also with nonvascular mortality, which is unexplained by its association with other known risk markers, or by the effects of blood pressure in middle- and old-age on both renal function and mortality. The results demonstrate the potential value of cystatin C as a biomarker for risk prediction, in addition to detection of early stages of renal impairment, in older people without established chronic kidney disease. The findings also highlight the need for further study of the determinants of renal function and their relevance for health in old age.

Conflicts of interest statement

None of the authors have any conflicts of interest in relation to this report.

Funding sources

The study was supported by grants from the British Heart Foundation and Medical Research Council, United Kingdom. JRE receives an Intermediate Research Fellowship from the British Heart Foundation.

Acknowledgements

We are grateful to all the participants in the Whitehall study of London Civil Servants. We acknowledge Martin Shipley, Astrid Fletcher, Elizabeth Breeze, Dave Leon, Michael Marmot, Martin Radley and Rory Collins for their help in carrying out the Whitehall study.

Ancillary