SECUENCIA DE ENSEÑANZA APRENDIZAJE 1 Extraer y Envolver

Twenty four hour average concentrations of the pollutants and the description of the meteorological data for the six cities for the study period are given in Table 10. PNC and PM2.5

were highest in the southern centres and lowest in Stockholm and Helsinki, while Augsburg showed intermediate levels (Figure 8) (Ruckerl et al., 2007), (Appendix III).

For hs-CRP, no associations between ambient air pollution and serum concentrations were observed for any of the analysed time lags (Table 11a). IL-6 and Fibrinogen on the other hand, which were examined in the same population in previous analyses (Ruckerl et al., 2007), (Appendix III), showed positive associations with levels of air pollution. In particular, an increase in IL-6 was seen when PNC were elevated 12–17 hours before the clinical visit and an increase in fibrinogen in association with a five day cumulative exposure to PM10 (Table 11b). Analyses of

effect modification showed that for fibrinogen associations remained for the five day average exposure to PM10 for patients with elevated HbA1c levels (Figure 9).

Table 10. 24-hour average concentrations of the ambient air pollutants and meteorological parameters from 6 European cities during the AIRGENE study period

Helsinki Stockholm Augsburg Rome Barcelona Athens

Study perioda 05.09.03 – 02.06.04 30.08.03 – 24.06.04 14.05.03 – 24.02.04 20.09.03 – 15.07.04 30.08.03 – 16.06.04 08.09.03 – 30.07.04

Mean 95th Mean 95th Mean 95th Mean 95th Mean 95th Mean 95th

PNC [1/cm³] 8534 15077 9748 17578 11876b _{25135 35450}b _{69226 18133}b _{36526 20589}b ₄₇₅₇₃ PM2.5 [µg/m³] 8.2 19.4 8.8 19.1 17.4 29.3 24.5b 54.1 24.2b 62.7 23.0b 46.0 PM10 [µg/m³] 17.1 36.1 17.8 40.3 33.1 56.6 42.1 76.0 40.7b 88.7 38.5 64.6 CO [mg/m³] 0.31 0.46 0.29 0.43 0.58 1.00 1.40 2.47 0.59 0.92 1.48 3.23 NO2 [µg/m³] 28.6 49.8 18.6 32.6 40.0 61.2 67.0 90.8 50.5 79.6 50.1 73.0 NO [µg/m³] 12.5 40.7 4.9 15.5 30.0 80.4 65.7 164.0 37.7 88.4 41.8 144.6 SO2 [µg/m³] 4.2 10.1 1.9 4.9 3.0 5.7 4.1 9.2 4.7 9.6 10.3 23.2 O3c [µg/m³] 46.8 89.0 60.6 96.9 54.4 115.3 45.3 99.6 28.2 76.5 59.8 100.2 Air temperature [°C] 3.1 14.7 4.7 15.1 10.2 25.1 13.4 23.9 15.2 23.2 17.6 29.3 Relative humidity [%] 76 91 82 94 69 92 80 95 67 86 67 84

a_{The study period started 5 days before the first measurement because air pollution concentrations up to 5 days before the blood withdrawals were considered.} b_{Data available on less than 95% of the days;} c_{Average of 8 maximum hourly values within the past 24 hours}

PNC: particle number concentration, PM2.5: mass concentration of particles less than 2.5 µm in diameter;PM10: mass concentration of particles less than 10 µm in diameter; CO: carbon monoxide; NO2: nitrogen dioxide; NO: nitrogen monoxide; SO2: sulphur dioxide; O3: ozone;

Figure 8. Time series of air pollution (PNC and PM2.5) and air temperature in the six European cities of

Table 11a: Effects of particulate air pollutants on hs-CRP per increase in interquartile range (IQR) of air pollutant

Table 11b: Effects of particulate air pollutants on IL-6 and Fibrinogen per increase in interquartile range (IQR) of air pollutant

Hs-CRPa (all cities)

Pollutant Time prior to

blood withdrawal IQR

%-change (GMb) 95% CI c p-value heterogeneityd Lower Upper PNCe _{Lag 0 11852 1.33 -3.05 5.90 0.047} Lag 1 11852 -1.52 -4.39 1.45 0.19 Lag 2 11852 -1.63 -6.70 3.71 0.019 5-d-aver. 11003 -0.08 -3.78 3.75 0.12 PM2.5f Lag 0 11.0 0.11 -1.95 2.21 0.71 Lag 1 11.0 -0.06 -1.98 1.90 0.70 Lag 2 11.0 0.11 -1.80 2.06 0.86 5-d-aver. 8.6 -0.13 -2.15 1.92 0.94 PM10g Lag 0 17.4 -0.71 -2.75 1.37 0.16 Lag 1 17.4 -0.63 -2.61 1.39 0.23 Lag 2 17.4 -1.42 -4.23 1.47 0.086 5-d-aver. 13.5 -1.35 -3.45 0.79 0.19 IL-6h (all cities) Fibrinogen (all cities except Athens)

Pollutant Time prior to blood withdrawal IQR %- change (GMb) 95% CI p-value hetero- geneityc %- change (AMi) 95% CI p-value hetero- geneity

Lower Upper Lower Upper

PNCe _{Lag 0 11852 1.88}† _{-0.16 3.97 0.72 0.40 -0.40 1.19 0.54} Lag 1 11852 -0.67 -2.56 1.25 0.64 0.11 -0.69 0.91 0.12 Lag 2 11852 -2.12† -4.03 -0.17 0.055 0.09 -0.71 0.90 0.045 5-d-aver. 11003 -0.93 -3.37 1.56 0.084 0.50 -2.20 3.20 0.009 PM2.5f Lag 0 11.0 0.46 -0.89 1.83 0.26 0.05 -0.48 0.58 0.36 Lag 1 11.0 -0.39 -1.69 0.93 0.70 0.17 -0.35 0.69 0.55 Lag 2 11.0 -0.23 -1.53 1.07 0.57 0.20 -0.32 0.71 0.26 5-d-aver. 8.6 0.05 -1.37 1.50 0.66 0.38 -0.21 0.96 0.21 PM10g Lag 0 17.4 -0.34 -1.66 0.99 0.45 0.06 -0.43 0.55 0.53 Lag 1 17.4 -0.69 -1.95 0.58 0.43 0.14 -0.35 0.63 0.83 Lag 2 17.4 -1.59 -3.99 0.88 0.0030 0.24 -0.24 0.72 0.25 5-d-aver. 13.5 -0.87 -2.28 0.55 0.15 0.60* 0.10 1.09 0.26

a_{CRP: c-reactive protein;} b_{GM: geometric mean, geometric mean is the antilog of arithmetic mean of log-transformed variable,} c_{CI: confidence interval,} d_{heterogeneity determined with}

χ²-test, α=10%; ePNC: particle number concentration; fPM2.5: mass

concentration of particles less than 2.5 µm in diameter; g_PM

10: mass concentration of particles less than 10 µm in diameter, h_{IL-6: interleukin 6,} i_{AM: arithmetic mean; *p<0.05;} †

> 6.5 <= 6.5 % c ha ng e in f ib ri no ge n pe r IQ R i nc re as e in P M10 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5

Figure 9. Effect modification of fibrinogen (five day average exposure) by glycosylised haemoglobin (HbA1c).

5 DISCUSSION

Data from the AIRGENE study, a large European multi-centre study on MI survivors, were used 1. a) to confirm known associations between time-invariant patient characteristics and hs-

CRP concentrations,

b) to extend the number of the characteristics published so far and

c) to study the influence of the same parameters on the variation of hs-CRP, hypothesising that the diagnosis of certain diseases would be related to higher hs-CRP concentrations and a more pronounced variation.

2. to examine the associations between hs-CRP and life-style parameters in the 24 hours before the blood draw, hypothesising that the consumption of tea and alcohol would lead to lower hs-CRP concentrations, while extreme stress or anger, active smoking and physical activity would show a positive association and

3. to examine the association of hs-CRP concentrations and environmental parameters, especially particulate air pollutants, assuming a positive linear relationship.

Hs-CRP values measured in the AIRGENE population were on average slightly higher than in other studies. Subsamples of the three population based MONICA/KORA studies from Augsburg, Glasgow and Lille, for example, showed median concentrations of between 1.30 mg/L (Glasgow and Lille) and 1.38 mg/L (Augsburg) (Imhof et al., 2004), while the median hs-CRP concentration in the whole AIRGENE population was 1.61 mg/L. Similar concentrations to those of the MONICA/KORA populations were also seen in about 4,000 middle aged men and women of the Swedisch Malmö Diet and cancer study (Rosvall et al., 2007). The slightly higher concentrations in these data might in part be explained by the fact that the AIRGENE population consisted of MI survivors, who are likely to have higher hs-CRP concentrations due to their disease history. Ridker et al. (Ridker et al., 1998d), for example, reported median concentrations of 3.1 mg/L in participants who subsequently had recurrent MI, while their age and sex matched

healthy controls had a median concentration of 2.8 mg/L. Moreover, Geffken et al. (2001) reported from a study of approximately 6,000 elderly US citicens, that people without clinical cardiovascular disease had significantly lower mean hs-CRP concentrations compared to participants with clinical cardiovascular disease.

Considering the different regions of the AIRGENE study, the results of Imhof et al. (2004) for Augsburg are almost the same as for the AIRGENE Augsburg subsample. Also, the comparatively low levels for the Swedish and the relatively high concentrations for the Spanish subpopoulations in AIRGENE were seen in other studies (Garcia-Lorda et al., 2006; Rosvall et al., 2007).

Some studies report mean values, rather than the median; however, these data more or less reflect the picture found for the median concentrations. Mean hs-CRP concentrations for the whole AIRGENE population was 2.59 mg/L and therefore higher than mean concentration in two large studies in healthy voluteers (Geffken et al., 2001; Sung, 2006). Population based data from the Greek ATTICA study (Panagiotakos et al., 2004) were again lower than the mean concentrations of the Greek subpopulation of AIRGENE, as were data from a Finnish study (Saltevo et al., 2007).

As hs-CRP usually shows a skewed distribution, values are often log-transformed for analyses and the antilog of arithmetic mean, the geometric mean, is reported. A log-transformation reduces skewness in the data (Bland et al., 1996). The geometric mean was 1.39 mg/L in the AIRGENE population. A comparison of geometric means from the MONICA studies in Augsburg, Glasgow and Lille with the geometric mean of the AIRGENE population showed that the geometric mean of MONICA Augsburg was lower than that of the whole AIRGENE population, while those of Glasgow and Lille were higher. Again, geometric means of patients with chronic cardiovascular disease were higher than for those without (Danesh et al., 2004; Hoffmeister et al., 2001; Tuomisto et al., 2006).

5.1 Determinants of hs-CRP concentration and variation

Results of the AIRGENE data mostly confirmed previously published associations between time- invariant patient characteristics and hs-CRP. In contrast to the initial hypothesis, the diagnosis of any of the diseases recorded in this study seemed not to be related to higher hs-CRP levels, with the exception of chronic bronchitis. Moreover, the variation of hs-CRP was significantly lower in patients with the studied co-morbidities, the only exceptions being the diagnosis of diabetes and a baseline level of HbA1c of at least 6.5%.

A variety of studies have examined influences on the level of hs-CRP and most of the results are in line with the results from the AIRGENE dataset, such as for elevated BMI and obesity (Garcia- Lorda et al., 2006; Greenfield et al., 2004; Sung, 2006; Thorand et al., 2006; Verdaet et al., 2004), for smokers compared to non-smokers (Frohlich et al., 2003; Garcia-Lorda et al., 2006; Geffken et al., 2001; Verdaet et al., 2004) and for subjects with low HDL-cholesterol levels (Garcia-Lorda et al., 2006; Greenfield et al., 2004).

While some investigators did not report any sex differences (Garcia-Lorda et al., 2006), others found lower concentrations in men consistent with the results from AIRGENE (Bowden et al., 2006; Frohlich et al., 2003; Hutchinson et al., 2000; Imhof et al., 2004). Hutchinson et al. (2000) hypothesised that the sex difference might be due to oestrogen intake in women and a study on diabetic women showed significantly higher hs-CRP concentrations in those receiving HRT (Bowden et al., 2006). In the AIRGENE data, intake of HRT showed a slightly positive but non- significant association with hs-CRP (data not shown), consistent with this hypothesis.

As for the influence of age on hs-CRP concentrations, some authors found a positive linear relationship (Hutchinson et al., 2000) while a lack of association has also been reported (Greenfield et al., 2004). It appears that a U-shaped function, as seen in these data, has not been reported before; this could be due the way the relationships were modelled and/or to the fact that the AIRGENE data were based on MI survivors, while most studies have been conducted in the general population.

A higher variation in diabetic patients compared to non-diabetics was seen, however geometric mean hs-CRP concentrations did not differ significantly, in contrast to results of other authors (Pradhan et al., 2001). Subjects with elevated HbA1c (≥ 6.5%), an indicator of glucose control,

however, showed higher hs-CRP concentrations and higher variation even in this comparatively homogeneous population of MI survivors. In this dataset, high levels of HbA1c seem to reflect uncontrolled rather than undiagnosed diabetes as 89% of those with HbA1c values above 6.5% reported a diagnosis of diabetes. Only half of the AIRGENE population diagnosed with diabetes on the other hand met the HbA1c criterion of 6.5%. This might indicate that metabolically stable diabetic patients are at less risk compared to unstable diabetics.

No publication seems to deal with the association between hs-CRP and N-terminal proB-type natriuretic peptide (Nt-pro BNP), an indicator for left ventricular dysfunction. However, in patients with chronic heart failure a worse New York Heart Association classification was associated with higher concentrations of hs-CRP and Nt-pro BNP (Windram et al., 2007). Moreover, a study in patients with non-ST elevation acute coronary syndromes showed that Troponin T, NT pro-BNP and hs-CRP were predictors of risk for major events such as death, new acute coronary syndrome, revascularization and heart failure in a multivariate analysis. The authors followed that a multimarker approach based on Troponin T, hs-CRP and NT-proBNP provides a more accurate prognosis regarding acute coronary syndrome, with a worse outcome for those with two or three elevated biomarkers (Tello-Montoliu et al., 2007). These studies indicate, that hs-CRP and Nt-pro BNP have a positive association, as found in the AIRGENE population.

None of the so far published works examines the within-patient variation of hs-CRP concentrations among different subgroups such as males and females or patients having certain medical conditions. Interestingly, the analysed data showed that an increase in geometric mean hs-CRP concentrations and a higher variation were not necessarily related. Subjects that reported angina pectoris, CHF or emphysema showed a lower variation compared to participants who did not report any of these disorders. These findings remained statistically significant even after

adjustment for multiple testing and associated medication intake. Emphysema is often caused by smoking (Petty, 2002) and in the AIRGENE population more than 80% of the emphysema patients were past or current smokers. As emphysema and early stage CHF do not necessarily include an inflammatory component, it is also conceivable that the lower variation of hs-CRP in these patients is merely a marker for a different mechanism, such as an underlying genetic component. Studies on twins have shown that there is a substantial genetic contribution to baseline hs-CRP concentrations (Pepys et al., 2003) and in genetical analyses of the Airgene dataset minor alleles of several variants in selected candidate genes such as CRP, Fibrinogen, IL- 6 and IL-10 were found to be significantly associated with intra-individual variation of hs-CRP

concentrations (Kolz et al., 2007).

Several studies, also in agreement with the data presented here, have shown that statin therapy reduces circulating hs-CRP (Albert et al., 2001a; Albert et al., 2001b; Libby et al., 2002; Ridker et al., 2001; Ridker et al., 2008; Rosenson et al., 2003). Clinical trials indicate that the proven efficacy of various compounds in primary and secondary prevention of coronary heart disease, such as statins and aspirin, may, at least in part, be mediated by modifying the consequences of the inflammatory response on vascular risk. In particular, additional non-lipid, anti- atherothrombotic and anti-inflammatory effects have been suggested for some statins and the clinical benefit in several studies occurred too early to be explained simply by an impact on the atherosclerotic process (Rosenson et al., 1998). In addition to a change in the lipid profile, especially on the LDL cholesterol level, and an improvement in endothelial function, some statins reduce inflammatory cells within the plaque. This leads to increased plaque stability as fewer enzymes are produced which degrade the extra-cellular matrix and weaken the plaque. Moreover, statins improve the haemostatic balance due to a reduction in platelet aggregation and various coagulation factors such as tissue factor (Rosenson et al., 1998; Welty et al., 1997). Data from the Cholesterol and Recurrent Events (CARE) trial (Ridker et al., 1998d) showed that the relative risk of recurrent coronary events was highest in the group with consistent evidence of inflammation, indicated by hs-CRP and SAA levels above the 90th percentile, which was assigned

to placebo, while the lowest risk was found in the group of patients without signs of inflammation who were additionally on statins. The results also demonstrated that statin intake in the group of patients with high hs-CRP and SAA levels reduced the risk of a recurrent coronary event by 54% to the level of those on placebo but without inflammatory activity. Moreover, the JUPITER (Justification for the Use of Statins in Prevention: an Intervention Trial Evaluating Rosuvastatin) trial showed for the first time in a prospective study a beneficial effect of statin intake on cardiovascular endpoints in addition to reduced CRP and LDL cholesterol levels in people with elevated hs-CRP but LDL cholesterol below current treatment levels (Ridker et al., 2008). Further evidence has been found in studies indicating a decrease of IL-6 during treatment with some statins (Koenig, 1999; Rosenson et al., 1999). A reduction in IL-6 might then lead to a reduced production of CRP.

In addition, CRP lowering effects have been seen with ASA (Ikonomidis et al., 1999), and ACE inhibitors (Soriano et al., 2007) however, in the AIRGENE data, associations were not significant. An increase in CRP in association with HRT has been reported (Cushman et al., 1999; Ridker et al., 2000) and could be confirmed in this dataset.

Cholesterol, which is transported in the form of lipoproteins, plays an important role in the development of atherosclerosis. LDL cholesterol serves as the principal carrier, and it provides an exogenous source of cholesterol and other cellular nutrients to hepatic and extra hepatic cells through LDL receptor-mediated uptake. Early stages of arterial lipoprotein modification, marked by generation of bioactive products of lipid peroxidation, can occur with little change in cellular receptor recognition. The uptake of these modified products facilitates the intracellular accumulation of lipoprotein-derived cholesterol and cholesteryl esters, characteristic of arterial foam cell formation, one of the first steps in the development of atherosclerotic lesions. HDL cholesterol, on the other hand, is a component of reverse cholesterol transport, which takes cholesterol back from extrahepatic tissues to the liver (Hajjar, 1997). Statin therapy has been shown to be associated with a regression of coronary artery disease, when LDL is substantially reduced and HDL is increased. This might indicate that statin benefits are derived from both

reductions in atherogenic lipoprotein levels and increases in HDL (Nicholls et al., 2007). Data of 770 males of the AIRGENE population showed strong positive associations with CRP for total cholesterol and LDL cholesterol but no association for HDL cholesterol when parameters were added separately to the base model. HDL and total cholesterol had opposing associations with a stronger effect of total cholesterol on hs-CRP. Therefore, the negative association between HDL and hs-CRP only became apparent after adjusting the model for total cholesterol. Adjustment for LDL, on the other hand, weakened the association for total cholesterol and hs-CRP, while LDL cholesterol only showed significantly positive associations without adjustment for total cholesterol. In previous studies, HDL was negatively associated with hs-CRP concentrations while neither LDL nor total cholesterol showed significant associations, however, the methods used in these studies were slightly different to the one used here (Fredrikson et al., 2004; Garcia- Lorda et al., 2006; Ryu et al., 2005).

Whether different factors affect each other and, if so, how, remains speculative. It is possible that a combination of variables amplifies the variation although it is also conceivable that certain combinations of factors may reduce variation. Additionally, factors that are associated with a high variation could just be indicators for a different mechanism. For example, the increase in variation that was seen in our data in association with medication intake might be a direct effect of the medication itself. However, it is more likely that the high variation is due to the underlying disease which led to the prescription of the drug.

5.2 Short-term influences on hs-CRP

Regarding time-varying life-style parameters, results point in the expected direction for alcohol and tea consumption and active smoking, however they were not significant, possibly due to a lack of patients with a variation in those endpoints.

In respect to tea and alcohol consumption, no publication seems to have addressed its effects within the 24 hours post-consumption. A slight decrease in hs-CRP in association with tea and alcohol intake was found; however the number of subjects that contribute to the effect was small, with only 260 patients reporting varying tea consumption and 500 reporting varying alcohol

intake. Several authors have shown a hs-CRP lowering effect of regular black tea (De Bacquer et al., 2006; Steptoe et al., 2007) and of moderate alcohol consumption (Imhof et al., 2004).

Several studies have demonstrated that regular moderate to high exercise leads to a decrease in hs-CRP concentrations (Elosua et al., 2005; Pitsavos et al., 2007) although results are conflicting (Fontana et al., 2007), and some authors assign the detected negative association to a lower BMI in those who exercise rather than to a direct effect of physical activity on inflammatory markers (Elosua et al., 2005; Verdaet et al., 2004). Short term influences, however, have rarely been looked at. Research in sports medicine has examined the effect of extremely strenuous activities on inflammatory markers, among them hs-CRP, and studies report a positive association (Kim et al., 2007; Margeli et al., 2005). However, the reported studies were conducted in people whose