Serum amyloid A is independently associated with metabolic risk factors but not with early atherosclerosis: the Cardiovascular Risk in Young Finns Study

Authors


Juulia Jylhävä, Department of Microbiology and Immunology, University of Tampere, Medical School, FIN-33014 Tampere, Finland.
(fax: +358 3 3551 6173; e-mail: juulia.jylhava@uta.fi).

Abstract.

Background.  Serum amyloid A (SAA) is a sensitive marker of inflammation and its elevation has been implicated in obesity and in cardiovascular disease, yet data on its regulation in young adults or on its role in early atherosclerosis is scarce. We investigated which factors explain the variation in SAA and analysed whether SAA could be associated with preclinical atherosclerosis.

Methods.  Serum amyloid A levels were measured in participants of the Cardiovascular Risk in Young Finns Study (n = 2280, n = 1254 women, n = 1026 men). Correlates and determinants of SAA were analysed and the effect of SAA on subclinical atherosclerosis, measured as intima-media thickness (IMT) and carotid artery compliance, was evaluated with risk-factor adjusted models.

Results.  Serum amyloid A correlated directly and independently of BMI with C-reactive protein (CRP), waist circumference and leptin in both sexes, with total cholesterol, LDL cholesterol and ApolipoproteinA1 (ApoA1) in women and with triglycerides, insulin levels and insulin resistance in men. Use of combined oral contraceptives and intrauterine device was also associated with SAA levels. Determinants for SAA included CRP, leptin and ApoA1 in women, and CRP, leptin and HDL cholesterol in men. SAA levels correlated with carotid compliance in both sexes and with IMT in men, yet SAA had no independent effect on IMT or carotid compliance in multivariable analysis.

Conclusions.  Serum amyloid A was associated with several metabolic risk factors but was not an independent predictor of IMT or carotid artery compliance. Further longitudinal studies will show whether SAA holds a prognostic value as a risk marker, analogously to CRP.

Introduction

Inflammation has been proposed to have an essential pathophysiological role in atherosclerosis, metabolic syndrome and diabetes [1, 2]. The biochemical measurements used to detect the low-grade inflammatory state in these disorders has mainly focused on C-reactive protein (CRP), which frequently – yet not consistently – has been associated with severity or poorer prognosis in these disorders [1–3]. Another acute phase reactant, serum amyloid A (SAA), has also proven to be a suitable and sensitive indicator of the inflammation involved in various stages of these diseases [3, 4]. However, it is yet to be resolved whether CRP and SAA can act as functional risk factors or if they are merely risk markers, i.e. indicators of the systemic nature of the low-grade inflammatory conditions. Nevertheless, elevation in SAA has been shown to predict cardiovascular events analogously with or even better than CRP [5–7] and in this sense, it has been speculated that SAA could be one of the links or even a proatherogenic risk factor between the inflammatory metabolic disorders and cardiovascular disease (CVD) [8, 9].

The human genome encompasses four SAA genes, of which three encode functional proteins [4]. SAA1 and SAA2 are highly homologous reactants whose concentration can increase up to 1000-fold upon infection or trauma, whereas SAA3 is a pseudogene and SAA4 is a constitutively expressed minor constituent of non-acute-phase HDL [10]. During the acute phase, liver is considered to be the major site of SAA synthesis, whilst in nonacute low inflammatory conditions, such as obesity, the role of adipocytes as the secretory site of SAA has recently been firmly established [11–13]. Expression of SAA is induced in response to proinflammatory cytokines such as IL-1β, IL-6 and TNF-α and circulating SAA associates predominantly with HDL particles, in which it displaces ApoA1 and thus alters the reverse cholesterol transport [10].

Because of its established correlation with CRP, leptin, body mass index (BMI) and other obesity indices [9, 11, 14, 15] as well as with cardiovascular events [5, 6], SAA has been a subject of several recent studies focusing on subjects with chronic diseases or obesity. However, little is known about its baseline distribution and regulation in younger healthy adults, likewise its association with early atherosclerosis. The aim of this study was therefore to establish the correlates and determinants of SAA in ostensibly healthy young adults as well as to evaluate the associations of SAA and CRP with early atherosclerosis.

Methods

Subjects

The study population consisted of participants of the Cardiovascular Risk in Young Finns study, which is an ongoing multicentre follow-up study on atherosclerosis risk factors in Finnish children and young adults. The first study was conducted in 1980, when the study was initiated and when the participants (n = 3596), who were randomly chosen from the national population registers of Helsinki, Tampere, Turku, Oulu, Kuopio and their rural surroundings, were 3, 6, 9, 12, 15 and 18 years of age. The study design has been described in more detail elsewhere [16, 17]. The follow-up, on which the data of the current study is based, was carried out in 2001, when the participants had reached 24, 27, 30, 33, 36 and 39 years of age. The parameters of early atherosclerosis, i.e. carotid artery compliance and intima-media thickness (IMT), as well as serum lipids, proteins and hormones, obesity indices, smoking habits, blood pressure values, alcohol consumption, physical activity, the presence of diabetes and rheumatic diseases were also recorded in this follow-up [18]. The study was approved by the local ethics committees and was conducted following the guidelines of the Declaration of Helsinki. All participants gave their written informed consent.

Clinical characteristics and biochemical measurements

Height and weight as well as waist and hip circumferences were measured and BMI and waist-hip ratio were calculated. Blood pressure was measured using a random zero sphygmomanometer and the mean of three measurements was used in the analyses. Venous blood samples for the determination of SAA, leptin, adiponectin, CRP, serum lipids (total cholesterol, HDL cholesterol, LDL cholesterol, triglycerides ApoA1 and ApoB), glucose and insulin were drawn after 12 h fasting. Serum SAA concentrations were measured with an ELISA kit with a detection limit of <0.004 mg L−1 (Human SAA, Biosource International, Camarillo, CA, USA). The inter-assay coefficient of variation (CV) was 10.6% at a mean level of 0.5988 mg L−1 and 12.6% at a mean level of 0.0613 mg L−1 and the intra-assay coefficient of variation was 5.8% at the mean level of 13.54 mg L−1. Serum CRP was determined using a high-sensitivity latex turbidometric immunoassay (Wako Chemicals GmbH, Neuss, Germany) with a detection limit of 0.06 mg L−1. The CV for repeated measurements was 3.3% at the mean level of 1.52 mg L−1 and 2.65% at the mean level of 2.52 mg L−1. Serum adiponectin was measured with radioimmunoassay (Human Adiponectin RIA kit; Linco Research, Inc., MO, USA) with an inter-assay CV of 11.9%. More detailed descriptions of the other analytical procedures and physical examination have been reported previously [18, 19]. Insulin resistance index was assessed with homeostasis assessment of insulin resistance (HOMA-IR), which was calculated based on fasting insulin and glucose values according to the formula: HOMA-IR = fasting glucose (mmol L−1) × fasting insulin (mU L−1)/22.5 [20]. Physical activity, alcohol consumption, smoking habits, use of combined oral contraceptives (COCs) or intrauterine device (IUD), rheumatic diseases, diabetes and recent infections were elicited by questionnaire [18, 19].

Physical activity was assessed as a metabolic equivalent (MET) index, in which one MET is the consumption of 1 kcal of a person per weight kilogram per hour in rest. MET was calculated from the product of intensity × frequency × duration and commuting physical activity. In the estimation of the physical activity during commuting to work place, the length of the journey and the means (i.e. whether it was travelled by foot or by bicycle) were taken into account. The coefficients for the variables were estimated from the previously established tables [21].

Subjects with diabetes (n = 22, SAA median 17.90 mg L−1), chronic rheumatic disease (n = 34, SAA 17.10 mg L−1), history of recent infection (n = 113, SAA 12.10 mg L−1) as well as pregnant women (n = 61, SAA 14.30 mg L−1) were excluded. Lactating women (n = 41) were not excluded as their plasma SAA (13.10 mg L−1) as well as CRP levels (0.7 mg L−1) were comparable to the study population (P = 0.090 for SAA and P = 0.849 for CRP). Subjects with triglycerides above 4 mmol L−1 (n = 30, SAA 12.95 mg L−1) were also excluded as the Friedwald formula used in LDL calculation could not be applied. In addition, we excluded women using COCs (n = 279, SAA 13.80 mg L−1) or levonorgestrel-releasing IUD (n = 104, SAA 8.70 mg L−1) from the analyses concerning the correlates, determinants and cardiovascular associations of SAA, as their SAA values deviated markedly (P < 0.001 and P = 0.03 respectively) from those of the nonusers. Rest of the excluded subjects (n = 128) were due to missing information in one or several measured variables. After subtracting these excluded individuals, we ended up with a population of n = 1509 subjects (n = 618 women and n = 891 men) for which we analysed the population data characteristics.

Carotid artery ultrasound measurements

Carotid ultrasound measurements were performed with a Sequoia 512 high-resolution ultrasound system (Acuson, CA, USA). Carotid artery compliance, which depicts the ability of the large arteries to expand under cardiac pulse pressure, was assessed from the formula ([DsDd]/Dd)/(PsPd), where Ds is the systolic diameter, Dd is the diastolic diameter, Ps is the systolic blood pressure and Pd is the diastolic blood pressure [22]. Mean IMT, a structural marker of vascular changes and a predictor of future cardiovascular events, was derived from a minimum of four measurements of the posterior wall of the left carotid artery (at approximately 10 mm proximal to the bifurcation) [22]. Both carotid compliance and IMT are continuous variables and they followed the normal distribution in our study population.

Statistical analyses

Statistical analyses were performed using spss version 15.0 (SPSS Inc., Chicago, IL, USA). All the analyses were carried out separately for men and women due to their markedly deviating SAA values (P < 0.001). Variables with skewed distribution (SAA, CRP, leptin, adiponectin, insulin and triglycerides) were log-transformed prior to analyses when normal distribution was required. Comparisons of the variables between sexes were performed with Student’s t-test, Mann–Whitney’s test and Chi-squared test as appropriate. In addition, Mann–Whitney’s test was used to test the heterogeneity of SAA levels between smokers and nonsmokers as well as between COC or IUD users and nonusers. The correlations between (log)SAA and clinical parameters as well as other serum markers were assessed with Pearson’s correlation coefficients, additionally adjusted for BMI. Determinants, i.e. the variables that explained the variation in SAA, were analysed with stepwise multivariable regression analysis in which all the variables that correlated significantly (P < 0.05) with SAA were included. Finally, multivariable linear regression analysis was used to evaluate whether SAA and CRP had an effect on preclinical atherosclerosis markers, carotid compliance and IMT. To exclude confounding effects, SAA and CRP were entered into the model as an added factor and the analysis was adjusted in a stepwise manner with the established risk-factors of atherosclerosis, i.e. age, BMI, LDL cholesterol, (log)triglycerides, glucose, systolic blood pressure (only IMT), daily smoking and physical activity. In all analyses, the level of P < 0.05 was considered statistically significant.

Results

Characteristics of the study population are shown in Table 1. The majority of the risk factors differed significantly between men and women, the only variables not deviating were age, physical activity index, insulin, insulin resistance index (HOMA-IR) and CRP (Table 1). A significant difference in SAA levels between men and women was observed (P < 0.001), but divergence in SAA levels was not, however, attenuated (data not shown) when the analysis was adjusted for leptin, adiponectin or ApoA1, which are the factors more elevated in women than in men.

Table 1.   Characteristics of the study population
VariableWomen (n = 618)Men (n = 891)P for difference
MeanSDMeanSD
  1. t-test for difference between groups.

  2. aMedian values and interquartile range (IQR), Mann–Whitney’s test for difference between groups.

  3. bChi-Squared test for difference between groups.

  4. ApoA1, apolipoproteinA1; ApoB, apolipoproteinB; BMI, body mass index; CRP, C-reactive protein; HOMA-IR, homeostasis assessment of insulin resistance; IMT, intima-media thickness; SAA, serum amyloid A.

Age (years)32.054.9431.665.020.134
BMI (kg m−2)24.354.6025.613.81<0.001
Waist circumference (cm)79.3911.4689.3010.55<0.001
Waist-hip ratio0.800.060.890.06<0.001
Systolic blood pressure (mmHg)115.5412.54128.9613.59<0.001
Diastolic blood pressure (mmHg)71.498.6974.839.09<0.001
Total cholesterol (mmol L−1)5.010.895.241.00<0.001
HDL cholesterol (mmol L−1)1.350.271.170.28<0.001
LDL cholesterol (mmol L−1)3.190.773.430.89<0.001
Triglycerides (mmol L−1)a0.900.70–1.201.200.90–1.80<0.001
ApoA1 (g L−1)1.500.221.410.21<0.001
ApoB (g L−1)0.970.231.120.26<0.001
Insulin (mU L−1)a6.004.00–8.006.004.00–9.000.403
Glucose (g L−1)4.910.415.150.41<0.001
HOMA-IR1.591.311.701.190.081
Leptin (ng mL−1)a12.597.58–19.504.072.40–6.46<0.001
Adiponectin (μg mL−1)a10.007.80–13.506.805.10–9.20<0.001
Homocysteine (μmol L−1)9.273.5710.793.89<0.001
CRP (mg L−)a0.560.26–1.360.580.29–1.340.679
SAA (mg L−1)a10.706.33–18.688.005.00–15.00<0.001
Physical activity index16.5914.7416.1617.400.649
Alcohol (drinks per week)3.726.448.7610.40<0.001
Smoking daily (% of total)b20 29 <0.001
Carotid compliance (%/10 mmHg)2.340.782.020.66<0.001
IMT (mm)0.580.090.590.100.003

Median SAA levels of daily smokers and nonsmokers did not differ amongst either sex (P = 0.546 for women and P = 0.642 for men). Women using COCs had significantly higher SAA values than nonusers [SAA median 13.60 mg L−1, (IQR 7.67–25.90) vs. SAA median 10.70 mg L−1, (IQR 6.33–18.68), P < 0.001], whereas women using levonorgestrel-releasing IUD had lower SAA values than nonusers [SAA median 8.75 mg L−1, (IQR 4.91–15.78) vs. 10.70 (IQR 6.33–18.68 mg L−1), P = 0.034]. To examine whether the difference in SAA levels could be attributed to the higher CRP concentration in COC using women [23], the analysis was adjusted for plasma CRP. This adjustment, however, did not abolish the observed differences.

SAA was found to correlate significantly (P < 0.05) and directly with BMI, waist circumference, waist-hip ratio, total cholesterol, LDL cholesterol, triglycerides, blood pressure, insulin, insulin resistance index (HOMA-IR), leptin, CRP and ApoB in both genders. However, significant and direct correlation was observed with glucose and physical activity index only in men, whilst a direct correlation with ApoA1 and adiponectin was observed only in women. Variables not correlating with SAA were age, glucose, use of alcohol, ApoB and homocysteine in women and use of alcohol, ApoA1 and homocysteine in men. All the significant correlations (P < 0.05) were further adjusted for BMI. Correlates that remained significant after BMI adjustment are presented in Table 2.

Table 2.   Correlates of clinical and biochemical parameters with (log)SAA in women and men
VariableWomen (n = 618)Adjusted for BMIMen (n = 891)Adjusted for BMI
raPraPraPraP
  1. The significant correlations were adjusted for BMI.

  2. aPearson’s correlation.

  3. ApoA1, apolipoproteinA1; ApoB, apolipoproteinB; BMI, body mass index; CRP, C-reactive protein; HOMA-IR, homeostasis assessment of insulin resistance; SAA, serum amyloid A.

Body mass index (BMI)0.328<0.001  0.281<0.001  
Waist circumference0.334<0.0010.0930.0210.312<0.0010.138<0.001
Waist-hip ratio0.214<0.0010.0340.4050.250<0.0010.0920.006
Age0.0250.732  0.0740.0280.0260.434
Total cholesterol0.189<0.0010.147<0.0010.0990.0030.0320.348
HDL cholesterol0.0560.163  −0.0660.0490.0100.776
LDL cholesterol0.144<0.0010.0940.0210.0710.0350.0100.766
(log)Triglycerides0.170<0.0010.0590.1440.180<0.0010.0770.022
ApoA10.121<0.0010.205<0.0010.0100.766  
ApoB0.219<0.0010.0660.1520.165<0.0010.0600.073
Systolic blood pressure0.163<0.0010.0480.2350.0800.017−0.0060.869
Diastolic blood pressure0.157<0.0010.0520.2020.1040.0020.0280.438
(log)Insulin0.183<0.0010.0140.7650.206<0.0010.0770.032
Glucose0.0220.593  0.0680.0420.0240.512
(log)Leptin0.390<0.0010.236<0.0010.342<0.0010.217<0.001
HOMA-IR0.191<0.0010.0020.9680.2000.0010.0750.025
(log)Adiponectin0.1000.016−0.0210.611−0.0510.131  
Homocysteine−0.0090.830  0.0230.492  
Physical activity index−0.0360.438  −0.0720.043−0.0680.058
Alcohol−0.0150.707  −0.0010.978  
(log)CRP0.650<0.0010.585<0.0010.703<0.0010.672<0.001

All variables correlating significantly (P < 0.05) with SAA were included in the multiple linear regression model. According to the stepwise multivariable linear regression analyses, the determinants for SAA in women were CRP (B = 0.479, P < 0.001), ApoA1 (B = 0.270, P < 0.001) and leptin (B = 0.121, P = 0.012) which together explained 44.9% of the total variation in SAA (Table 3a). In men, the determinants for SAA were CRP (B = 0.610, P < 0.001), leptin (B = 0.112, P = 0.007) and HDL cholesterol (B = 0.097, P = 0.017), together explaining 49.7% of the variation in SAA (Table 3a). However, due to the strong intercorrelation and similar proinflammatory regulation of SAA and CRP we also wished to analyze the determinants of SAA without the effect of CRP. In this model, the determinants for SAA in women were leptin (B = 0.424, P < 0.001), ApoA1 (B = 0.270, P < 0.001) and body mass index (B = 0.011, P = 0.013) and in men they were leptin (B = 0.340, P < 0.001) and waist circumference (B = 0.001, P = 0.006) (Table 3b). These models, however, explained only 18.5% (in women) and 12.6% (in men) of the variation in SAA.

Table 3.   Determinants for serum amyloid A (SAA) in adjusted multiple linear regression models in women and in men with (a) and without (b) the effect of C-reactive protein (CRP)
VariableWomen (n = 618) Men (n = 891)
BSEPBSEP
  1. Variables correlating significantly with (log)SAA (P < 0.05) were entered into the model in a stepwise manner.

  2. Variables entered for women were BMI, waist circumference, waist-hip ratio, total cholesterol, LDL cholesterol, (log)triglycerides, systolic blood pressure, diastolic blood pressure, (log)adiponectin, (log)insulin, glucose, insulin resistance (HOMA-IR), (log)leptin, (log)CRP (only in model a), ApoA1 and ApoB. Variables entered for men were BMI, waist circumference, waist-hip ratio, age, total cholesterol, HDL cholesterol, LDL cholesterol, (log)triglycerides, systolic blood pressure, diastolic blood pressure, (log)insulin, glucose, insulin resistance (HOMA-IR), (log)leptin, physical activity index, (log)CRP (only in model a) and ApoB.

  3. ApoA1, apolipoproteinA1; BMI, body mass index; CRP, C-reactive protein.

(a)
 (log)CRP 0.479±0.028<0.0010.610±0.025<0.001
  (log)Leptin0.121±0.0480.0120.112±0.0410.007
 ApoA10.270±0.054<0.001   
 HDL cholesterol    0.097±0.0400.017
 R2 = 0.449R2 = 0.497
(b)
  (log)Leptin0.424±0.070<0.0010.340±0.071<0.001
 ApoA10.270±0.054<0.001   
 BMI0.011±0.0040.013   
 Waist circumference    0.001±0.0010.006
 R2 = 0.185R2 = 0.126

Stepwise multiple linear regression analyses were used to assess the effect of SAA and CRP on the early atherosclerosis markers carotid compliance and IMT. The effect of SAA and CRP was found to be very similar, yet dependent on other risk factors, mainly BMI and serum lipids (Table 4). In univariable analysis (model 1, Table 4) both SAA and CRP correlated inversely with carotid compliance in both sexes and directly with IMT in men (Table 4). Neither SAA, nor CRP correlated with IMT in females (Table 4). A similar trend was also observable after adjusting for age, smoking and physical activity (model 2 in Table 4). However, after adjusting for BMI, (log)triglycerides, glucose, LDL cholesterol and systolic blood pressure (only for IMT), the effects of SAA and CRP on carotid compliance and IMT were attenuated to the null (model 3 in Table 4). Only the association of CRP with carotid compliance in males remained of borderline significance (model 3 in Table 4).

Table 4.   Serum amyloid A (SAA) and C-reactive protein (CRP) as markers of IMT and carotid compliance in multivariable stepwise regression model
  Model 1Model 2Model 3
nBSEPnBSEPnBSEP
  1. Model 1. Only (log)SAA or (log)CRP as added factor.

  2. Model 2. The effect of (log)SAA or (log)CRP adjusted for age, physical activity and smoking.

  3. Model 3. The effect of (log)SAA or (log)CRP adjusted for age, physical activity, smoking, BMI, LDL cholesterol, glucose, systolic blood pressure (only for IMT) and (log)triglycerides.

  4. BMI, body mass index; IMT, intima-media thickness.

(log)SAA
 WomenIMT609−0.0010.0090.8814660.0020.0100.833466−0.0080.0100.427
Carotid compliance609−0.2060.0810.011466−0.2270.0920.014466−0.1000.0940.289
 MenIMT8830.0220.0080.0057730.0150.0080.0657710.0040.0080.620
Carotid compliance880−0.1700.0510.001771−0.1490.0520.004771−0.0710.0530.177
(log)CRP
 WomenIMT6090.0020.0070.7974660.0030.0080.740466−0.0120.0090.172
Carotid compliance609−0.1420.0640.026466−0.1630.0720.0234660.0140.0810.861
 MenIMT8830.0270.007<0.0017730.0210.0070.0037710.0080.0080.275
Carotid compliance880−0.2150.045<0.001771−0.1990.046<0.001771−0.1000.0500.045

Discussion

The results of this large cross-sectional study demonstrate as a novel finding that amongst ostensibly healthy adults, SAA levels correlate directly with several metabolic risk factors in women and in men, all correlations being independent of BMI. CRP was found to be the main determinant for SAA in both sexes, yet the additional model, in which we assessed the determinants for SAA without CRP, indicated that we might actually lack the best biological regulator(s) of SAA, as the additional model did not explain the variation on SAA very well.

Our results concerning the correlation of SAA with obesity indices, leptin and CRP corroborate earlier reports, which have shown that SAA levels correlate directly with CRP, leptin, obesity indices, body fat percentage and adipocyte size [9, 11, 14, 15]. However, others have not consistently documented the BMI-independent correlations between SAA and triglycerides, ApoA1 and HDL cholesterol [11, 13, 15]. These divergences may be due to the fact that the other studies have generally been performed on older and obese individuals in smaller study populations, in which SAA levels have also fallen drastically, concomitantly with weight loss. The BMI-independent correlations observed here therefore demonstrate that SAA can be regarded as an indicator of the metabolic status in lean and younger individuals as well. Along these lines, as all the determinants for SAA in our cohort, i.e. CRP, leptin and ApoA1/HDL cholesterol are related to lipid metabolism, we speculate that adipose tissue, regardless of its extent, contributes to SAA regulation via these factors. This hypothesis is plausible in light of the knowledge that excess/dysfunctional adipose tissue is a source of several pro-inflammatory reactants and that SAA and CRP are upregulated by pro-inflammatory stimuli, such as IL-1β, TNF-α and IL-6 [1, 4]. On the other hand, SAA, CRP and leptin can also induce pro-inflammatory cytokines [1, 9, 24], suggesting a positive feedback mechanism and a complex interconnection between these three factors. In addition, SAA has been shown to have a long-term effect in stimulating basal lipolysis [9] thus perhaps contributing to insulin resistance via increased release of free fatty acids into circulation. Our results corroborate this finding as we observed that in male subjects SAA correlates directly with insulin resistance index and triglycerides, independently of BMI. However, the molecular mechanisms interconnecting obesity and inflammation are still rather unknown and our data do not permit any conclusions on the direction or causality of this regulation scheme.

We observed that women had significantly elevated SAA levels compared with men; Lappalainen et al. (2008) also observed a similar sex difference amongst obese subjects [11]. Concurrent observations have also been reported for SAA mRNA production in adipose tissue [11, 13], though the reason for the gender difference is not known. We sought to explain the gender difference by the higher leptin and ApoA1 concentrations in women, but adjustment with these factors did not change the result, indicating that the divergence is related to other factors, possibly body composition or hormonal profile. Unfortunately, we have no data available on body fat percentage or sex hormones in our study population.

Nevertheless, hormonal contribution to SAA regulation has been reported in premenopausal women using COCs and also in postmenopausal women receiving oestrogen replacement therapy (ERT); SAA levels were significantly elevated amongst those receiving oral-conjugated estrogens [25, 26]. In addition, it has recently been demonstrated that ERT-associated elevation in SAA was counteracted by oral administration of an androgenic progestin, medroxyprogesterone acetate [27]. In this study, we observed that premenopausal women using COCs had significantly higher SAA levels compared with nonusers, whereas women using levonorgestrel-releasing IUD had lower median SAA than nonusers. Although not directly comparable, our results lend support to these observations on hormonal regulation of SAA. First, the higher SAA levels amongst the COC using women were not merely secondary to their higher CRP concentrations [23] as the difference remained unchanged after adjusting for plasma CRP. Second, the IUD used by our study women contained levonorgestrel which is an androgenic, testosterone-derived progestin [28], thus likely to excrete anti-inflammatory effects analogous to medroxyprogesterone acetate. The mode of action of these hormonal components most probably deals with the first-pass hepatic effect, as transdermal ERT has an opposite effect on SAA levels compared with oral ERT [25]. However, whether systemic inflammatory reaction is involved in this regulation of SAA is still somewhat uncertain [25, 26] and we were unfortunately unable to evaluate the matter as we did not measure the IL-6 or TNF-α levels in our cohort.

Whilst acute-phase SAA is almost entirely of liver origin, mounting evidence implies that adipose tissue is the main source of circulating SAA during the nonacute conditions, [9, 11–13, 15]. Also, endothelial cells, smooth muscle cells, monocytes and macrophages in atherosclerotic lesions have been reported to account for the extrahepatic production of SAA, as the presence of both SAA mRNA and protein products has been detected in these cell types [29–31]. Moreover, SAA is able to alter vascular proteoglycans in a proatherogenic manner [32] and to stimulate the production of various inflammatory mediators, such as TNF-α, IL-1β, IL-8, plasminogen activator inhibitor-1 and tissue factor in cultured vascular endothelial cells, neutrophils and monocytes [9, 24, 33]. In addition, activated neutrophils can induce formation of SAA-LDL complexes via lipoprotein oxidation in vitro [34]. Elevation in circulating SAA has also been related to acute cardiovascular events, with a better or equal prognostic value compared with CRP [5–7]. It therefore follows that, as CRP is already a widely used marker in prognosis of CVD [3, 7], these observations have raised the possibility that SAA could also be a proatherogenic risk factor and not merely a marker of systemic or local inflammation. However, data are still lacking with regard to the clinical relevance of minor elevation in SAA and also the cut-off value for low-grade elevation in SAA has not been established yet.

Our results, however, indicate that the association of SAA, as well as that of CRP, on the early atherosclerosis is mediated through BMI and serum lipids as the associations with carotid compliance and IMT were attenuated to the null when the multivariable model was adjusted for BMI and serum lipids. These results are in accordance with those of Wohlin et al. (2007) who observed no independent association between SAA and IMT in older men [35], yet Schillinger et al., (2005) reported that both SAA and CRP are associated with the progression of active but initially asymptomatic atherosclerosis in both sexes [36]. Therefore, functional studies are required to establish the role, if any, of SAA in atherogenesis, both in conditions conferring a risk of atherosclerosis per se– such as obesity – as well as in healthy and lean individuals. Although SAA is not an independent determinant for subclinical atherosclerosis in healthy young adults, we suggest that it is a marker of metabolic status and it remains to be discovered whether it contributes to the pathogenesis of early atherosclerosis via factors related to lipid metabolism. In conclusion, the results of this study demonstrate that SAA is associated with several metabolic risk factors in both genders regardless of BMI, and also with the use of COCs or IUD in premenopausal women, independently of CRP.

Sources of funding

This study was financially supported by the Emil Aaltonen Foundation (T.L.), the Tampere Tuberculosis Foundation, Competitive research funding of Pirkanmaa Hospital District, Turku University Central, Hospital Medical Fund, The Academy of Finland (grants 77841, 34316 and 210283), the Finnish Foundation of Cardiovascular Research, Juho Vainio Foundation and Yrjö Jahnsson Foundation.

Conflict of interest statement

None declared.

Acknowledgements

The authors wish to thank Sinikka Repo-Koskinen and Nina Peltonen for their skillful technical assistance.

Ancillary