C-reactive protein levels, haplotypes, and the risk of incident chronic obstructive pulmonary disease.
ABSTRACT Chronic obstructive pulmonary disease (COPD) is characterized by substantial chronic inflammation in the pulmonary compartment as well as in the systemic circulation.
To investigate potentially causal association, we examined whether serum levels of high-sensitivity C-reactive protein (hsCRP) and variations in the CRP gene are associated with the risk of developing COPD.
This study is part of the Rotterdam Study, a prospective population-based cohort study among subjects aged 55 years or older. At baseline, 6,836 subjects without COPD had a blood sample available for assessment of hsCRP serum levels and haplotypes of the CRP gene. We analyzed the association between hsCRP levels, CRP gene haplotypes, and incident COPD with Cox proportional hazard models, adjusted for age, sex, and other confounders.
High levels of hsCRP (>3 mg/L) were associated with a significantly increased risk of incident COPD (hazard ratio [HR], 1.7; 95% confidence interval [CI], 1.16-2.49) compared with persons with low levels (<1 mg/L). The risk remained increased after adjusting for potential confounders and introducing a latency period of 3 years. The risk was most pronounced in former smokers (HR, 2.2; 95% CI, 1.12-3.74). hsCRP was not a risk factor in never smokers. No CRP single nucleotide polymorphism or haplotype was associated with a significantly increased or decreased COPD risk.
Increased hsCRP levels are predictive for the occurrence of COPD in smokers. However, haplotypes of the CRP gene, which influence hsCRP levels, are not associated with an altered risk of developing COPD.
[show abstract] [hide abstract]
ABSTRACT: Background: Patients with chronic obstructive pulmonary disease (COPD) have elevated serum levels of ultrasensitive C reactive protein (CRPus). This raise may be related directly to COPD and its associated systemic inflammation or secondary to other factors such as smoking status, disease severity, acute exacerbations, or associated complications. Aim: To evaluate the potential causes of raised levels of CRPus in stable COPD patients. Patients and Methods: Cohorts of 133 mild-to-very severe COPD patients (41 current smokers), 31 never-smokers, and 33 current smoker controls were compared. Clinical assessments included body mass index (BMI), fat (FM) and fat-free mass (FFM) measurement by DEXA, forced expiratory volume in one second (FEV1), arterial oxygen tension (PaO2), six-minute walking test (SMWT), emphysema (EMPH) and right thigh muscle cross-sectional area (TMCSA), both quantified by high resolution computed tomography. Results: Serum CRPus levels were significantly higher in COPD patients than in controls (7 ± 4.2 and 3.7 ± 2.7 mg/L respectively; p < 0.0001). Being smoker did not influence CRPus levels. These levels were significantly correlated with FM (r = 0.30), BMI (r = 0.21), FEV1 (r = -0.21), number of acute exacerbations of the disease in the last year (r = 0.28), and PaO2 (r = -0.27). Using multivariate analysis FM, PaO2, and number of acute exacerbations of the disease in the last year had the strongest association with CRPus levels. Conclusions: CRPus is elevated in COPD patients, independent of smoking status. It is weakly associated with fat mass, arterial oxygen tension and frequency of exacerbations.Revista medica de Chile 05/2012; 140(5):569-78. · 0.33 Impact Factor
[show abstract] [hide abstract]
ABSTRACT: The Rotterdam Study is a prospective cohort study ongoing since 1990 in the city of Rotterdam in The Netherlands. The study targets cardiovascular, endocrine, hepatic, neurological, ophthalmic, psychiatric, dermatological, oncological, and respiratory diseases. As of 2008, 14,926 subjects aged 45 years or over comprise the Rotterdam Study cohort. The findings of the Rotterdam Study have been presented in over a 1,000 research articles and reports (see www.erasmus-epidemiology.nl/rotterdamstudy ). This article gives the rationale of the study and its design. It also presents a summary of the major findings and an update of the objectives and methods.European Journal of Epidemiology 08/2011; 26(8):657-86. · 4.71 Impact Factor
[show abstract] [hide abstract]
ABSTRACT: Chronic obstructive pulmonary disease, metabolic syndrome and diabetes mellitus are common and underdiagnosed medical conditions. It was predicted that chronic obstructive pulmonary disease will be the third leading cause of death worldwide by 2020. The healthcare burden of this disease is even greater if we consider the significant impact of chronic obstructive pulmonary disease on the cardiovascular morbidity and mortality.Chronic obstructive pulmonary disease may be considered as a novel risk factor for new onset type 2 diabetes mellitus via multiple pathophysiological alterations such as: inflammation and oxidative stress, insulin resistance, weight gain and alterations in metabolism of adipokines.On the other hand, diabetes may act as an independent factor, negatively affecting pulmonary structure and function. Diabetes is associated with an increased risk of pulmonary infections, disease exacerbations and worsened COPD outcomes. On the top of that, coexistent OSA may increase the risk for type 2 DM in some individuals.The current scientific data necessitate a greater outlook on chronic obstructive pulmonary disease and chronic obstructive pulmonary disease may be viewed as a risk factor for the new onset type 2 diabetes mellitus. Conversely, both types of diabetes mellitus should be viewed as strong contributing factors for the development of obstructive lung disease. Such approach can potentially improve the outcomes and medical control for both conditions, and, thus, decrease the healthcare burden of these major medical problems.Cardiovascular Diabetology 10/2012; · 3.35 Impact Factor
C-Reactive Protein Levels, Haplotypes and the Risk of
Incident Chronic Obstructive Pulmonary Disease
Yannick M.T.A. van Durme1,2; Katia M.C. Verhamme3; Albert-Jan. L.H.J. Aarnoudse2; Geert
R. Van Pottelberge1; Albert Hofman2,5; Jacqueline C.M. Witteman2,5; Guy F. Joos1; Guy G.
Brusselle1,2; Bruno H.Ch. Stricker2,4,5.
1. Department of Respiratory Medicine, Ghent University and Ghent University
Hospital, B-9000 Ghent, Belgium.
2. Department of Epidemiology, Erasmus University Medical Center, Rotterdam, the
3. Department of Medical Informatics, Erasmus University Medical Center, Rotterdam,
4. Inspectorate of Healthcare, The Hague, the Netherlands.
5. Members of the Netherlands Consortium on Healthy Aging (NCHA)
Correspondence and reprint requests:
Bruno H.Ch. Stricker, MB, PhD; Department of Epidemiology & Biostatistics, Erasmus
University Medical Center; PO Box 2040, 3000 DR Rotterdam, the Netherlands
Telephone number: 0031 10 408 7489; Fax: 0031 70 340 6793
The Rotterdam Study is supported by Erasmus Medical Center Rotterdam; the Erasmus
University Rotterdam; the Netherlands Organization for Scientific Research; the Netherlands
Organization for Health Research and Development; the Research Institute for Diseases in the
Elderly; the Ministry of Education, Culture and Science; and the Ministry of Health, Welfare
and Sports. This study was supported by the Netherlands Organization for Scientific Research
(NWO) grants 904-61-093 and 918-46-615. This study was supported by the Netherlands
Genomics Initiative (NGI) / Netherlands Organisation for Scientific Research (NOW) project
nr. 050-060-810 (NGI/NOW-NCHA). YvD received a travel grant by the Belgian Thoracic
Society and is a doctoral research fellow of the Fund for Scientific Research Flanders (FWO
Running Title: CRP levels, CRP haplotypes and COPD
This article has an online data supplement, which is accessible from this issue’s table of
content online at www.atsjournals.org
Word count: 3387
Page 1 of 39
AJRCCM Articles in Press. Published on December 18, 2008 as doi:10.1164/rccm.200810-1540OC
Copyright (C) 2008 by the American Thoracic Society.
Chronic obstructive pulmonary disease (COPD) is characterized by substantial chronic
inflammation in the pulmonary compartment as well as in the systemic circulation.
In order to investigate potentially causal association, we examined if serum levels of high-
sensitivity C-reactive protein (hsCRP) and variation in the CRP gene are associated with the
risk of developing COPD.
This study is part of the Rotterdam Study, a prospective population-based cohort study among
subjects ≥ 55 years. At baseline, 6836 subjects without COPD had a blood sample available
for assessment of hsCRP serum levels and haplotypes of the CRP gene. We analyzed the
association between hsCRP levels, CRP gene haplotypes and incident COPD with Cox
proportional hazard models, adjusted for age, gender and other confounders.
High levels of hsCRP (>3 mg/l) were associated with a significantly increased risk of incident
COPD (hazard ratio [HR], 1.7; 95% confidence interval [CI], 1.16 - 2.49) compared with
persons with low levels (< 1 mg/l). The risk remained increased after adjustment for potential
confounders and introducing a latency period of 3 years. The risk was most pronounced in
former smokers (HR, 2.2; 95% CI, 1.12 - 3.74). hsCRP was not a risk factor in never smokers.
No CRP single nucleotide polymorphism or haplotype was associated with a significantly
increased or decreased COPD risk.
Page 2 of 39
Increased hsCRP levels are predictive for the occurrence of COPD in smokers. However,
haplotypes of the CRP gene, which influence hsCRP levels, are not associated with an altered
risk of developing COPD.
Word count: 250
Key words: Chronic obstructive pulmonary disease, C-reactive protein, genetics,
Page 3 of 39
Chronic obstructive pulmonary disease (COPD) is a leading and still growing cause of
morbidity and mortality worldwide. Before 2020, COPD is expected to move from the sixth
to the third most common cause of death worldwide, whilst rising from fourth to third in
terms of morbidity (1). The disease is characterised by airflow limitation that is not fully
reversible. This airflow limitation is usually progressive and associated with an abnormal
inflammatory response of the lungs to noxious particles or gases (2). Smoking is a major
etiologic factor for COPD in the Western world (3). Interestingly, only 15 to 25% of cigarette
smokers seem to develop the disease, suggesting the presence of susceptibility genes for
COPD (4, 5).
There is increasing evidence of nonpulmonary effects in patients with COPD,
irrespective of treatment, resulting in deconditioning, osteoporosis, skeletal muscle
dysfunction, diabetes, cachexia, anaemia, anxiety/depression, lung cancer, cardiovascular
morbidity and mortality (6-9). Previous studies have shown that circulating markers of
systemic inflammation, such as interleukin-6 (IL-6) and C-reactive protein (CRP), are
elevated in patients with stable COPD and further increase during COPD exacerbations (10),
which may be a key link to most COPD-related co-morbidities or systemic consequences.
CRP is a well validated predictor of future myocardial infarction and stroke (11). In addition,
serum CRP levels seem to predict exacerbations and hospitalizations in patients with
established COPD (12). The impact of these systemic comorbidities and/or complications of
COPD is important. Smoking cessation and, as recently suggested by Celli et al,
pharmacotherapy with combined salmeterol plus fluticasone propionate (13), seem to reduce
the rate of decline in patients with COPD. Unfortunately, at this moment no treatment has a
substantial effect on the onset of the systemic consequences of COPD. New insights into the
mechanisms involved in the systemic complications of COPD are needed.
Page 4 of 39
CRP is an important component of the innate immune response, synthesized
predominantly in the liver in response to IL-6 (14, 15), but is also produced in the respiratory
epithelium during infection (16). CRP serum levels have major genetic determinants and
several single nucleotide polymorphisms (SNPs) in the CRP gene have already been shown to
be associated with differences in baseline CRP levels in human populations (17, 18). As a
result, populations have been studied for associations between CRP gene polymorphisms and
several diseases such as cancer (19), diabetes (20) and cardiovascular disease (18, 21-23). It is
not known whether variation in the CRP gene has an influence on the risk of developing
We hypothesised that serum levels of hsCRP would be an independent predictor of
incident COPD and that variation in the CRP gene would be associated with the susceptibility
of an individual to develop the disease. We studied this hypothesis in a large prospective
population-based cohort study with 15 years of follow-up. Some of the results of this study
have been previously reported in the form of an abstract (24, 25).
Page 5 of 39
Study population and baseline data collection
The present study is part of the Rotterdam Study, a population-based cohort study aimed at
assessing the occurrence of and risk factors for chronic diseases in the elderly. Objectives and
methods of the Rotterdam Study have been described elsewhere (26, 27). In short, the
Rotterdam study cohort includes 7983 participants aged ≥ 55 years, living in Ommoord, a
well-defined suburb of the city of Rotterdam, the Netherlands. Almost all participants
(99.8%) are of Caucasian descent. Baseline data were collected from 1989 until 1993.
Participants were visited at home at the start of the study for a standardized interview on their
health status. Since the start of the Rotterdam Study, each participant visited the research
center every 2 to 3 years, during which several measurements were performed. In addition,
participants were continuously monitored for the onset of major events which occured during
follow-up through automated linkage with files from general practitioners. Trained research
assistants collected information from medical records of the general practitioners, medical
specialists (e.g. respiratory physicians), hospitals and nursing homes. The medical ethics
committee of the Erasmus Medical Center, Rotterdam, approved the study. Participants gave
written informed consent, and permission to retrieve information from treating physicians.
Patient identification and validation
Information about the cohort definition has been previously described (27) and is in
detail provided in the online data supplement.
Page 6 of 39
In short, the diagnosis of COPD was classified as definite or probable. Definite COPD
was defined by a moderate-to-severe obstructive spirometry (FEV1/FVC < 0.7 and FEV1 <
80% predicted), and/or as COPD diagnosed by a specialist in internal medicine (mainly
respiratory physicians or internists with a subspeciality in respiratory medicine) based upon
the combination of clinical history, physical examination and spirometry. Probable COPD
was defined by a mild obstructive spirometry (FEV1/FVC < 0.7 and FEV1 ≥ 80% predicted)
and/or as COPD diagnosed by a physician in another medical speciality (e.g. a general
The index date was defined as the date of diagnosis of COPD found in the medical
reports, or the date of a first prescription of COPD medication, or the date of the obstructive
lung function examination, whichever came first. Follow-up time was defined as the time
period between cohort entry and the diagnosis of COPD, death, loss to follow-up, or the end
of the study period on 31 December 2004, whichever came first.
High-sensitivity CRP level measurement and CRP genotyping
Details on the methods for making these measurements are provided in the online data
supplement. Serum hsCRP levels and CRP genotypes were measured, as previously described
Firstly, we analyzed the association between serum hsCRP levels and time to incident COPD
with a Cox proportional hazard model. All analyses were done for the three categories of
exposure of hsCRP according to recommendations of the American Heart Association (28)
and for hsCRP levels as a continuous variable.
Page 7 of 39
Because of the skewed absolute values of CRP in our cohort, the natural logarithmic
transformation of CRP was entered as a continuous variable in the model. In the first model,
we adjusted for age and gender. A final model adjusted for all potential covariates that
changed the point estimate by more than 10%, or covariates that were independent risk factors
of the outcome according to previous literature (online data supplement). The proportional
hazards assumption was tested by drawing log minus log plots of the survival function.
Multiplicative interaction terms, e.g. (smoking behaviour)*(hsCRP), and analyses stratified
by gender and smoking behaviour, were introduced to explore for effect modification. We
also introduced a latency period of 3 years in which we restricted the analyses to those
subjects who were free of COPD for at least 3 years after the baseline CRP assessment.
Secondly, we studied the association between CRP SNPs and incident COPD. Hardy
Weinberg Equilibrium was calculated with a χ2 statistic. Cox proportional hazard models were
used to study the association of the three CRP SNPs, the four CRP haplotypes and incident
COPD, adjusted for age, gender and other confounders. Haplotype analyses were performed
by entering a 0-1-2 variable for haplotypes 2, 3 and 4 in the models, thus using haplotype 1,
the haplotype that was associated with the lowest serum CRP levels (18), as the reference
haplotype. We entered these variables both as continuous (haplotype dose-effect model) and
as categorical variables. Stratified analyses were performed by gender and smoking status
(never, current, former) to explore for effect modification. Trends were calculated by
including categorical variables as a continuous value in the Cox model.
The significance of the Cox proportional hazard analyses was interpreted by
calculating the 95% confidence intervals for each of the parameters of interest. For the trend
analyses, P-values below the conventional level of significance (P < 0.05) were considered as
statistically significant. All statistical analyses were performed using SPSS for Windows
version 15 (SPSS Inc, Chicago, IL).
Page 8 of 39
There were 928 well-defined COPD cases in the Rotterdam Study (11.6% of the total study
population). A total of 280 prevalent COPD cases were identified at baseline. In the
remaining study population, 648 participants developed COPD. This resulted in an overall
incidence rate of 9.2/1000 person years (95% confidence interval [CI], 8.5 - 10.0) (27). Of
the 7983 participants, 7081 donated a blood sample at baseline. Of these 7081 participants,
245 had prevalent COPD, resulting in a final study population of 6836 subjects with available
blood sample and without prevalent COPD (Figure 1). For logistic reasons, CRP
measurements were not determined in 417 participants. In addition, we excluded 317
participants with a CRP level above 10 mg/l and 78 subjects with baseline systemic
corticosteroid use, giving a total of 6024 participants to study the association between
baseline serum hsCRP levels and incident COPD (318 definite and 230 probable COPD
cases). Genotyping for all three tagging SNPs of the CRP gene was performed successfully in
5718 of the 6836 participants (316 definite and 213 probable COPD cases) (Figure 1). There
were no differences in the baseline characteristics between these two subgroups (Table E1
online data supplement).
Table 1 shows the baseline characteristics of the total study population (N = 6836).
The median age of the total study population when entering the Rotterdam Study was 68.6
years (interquartile range [IQR], 1.4 years) and 39.4 % of the participants were men. The
median follow-up time of the study cohort was 11.7 years (IQR, 6.49).
Page 9 of 39
High-sensitivity CRP levels
The median hsCRP serum level was 1.8 mg/l (IQR, 2.4) (Table 1). The CRP level distribution
was highly skewed to the left. After log-transformation, the residuals were normally
distributed (data not shown).
The increase in risk of developing COPD was most pronounced in the definite COPD
group for the highest category of hsCRP (> 3 mg/l) with a hazard ratio of 1.8 (95% CI, 1.32 –
2.46) in the first model adjusted for age and gender, and 1.7 (95% CI, 1.16 - 2.49) in the final
model with adjustment for all potential confounders (Table 2). The increase in risk was less
pronounced for the total (definite and probable) COPD group. The risk remained increased
after introduction of a 3-year latency period with a hazard of 1.5 (95% CI, 0.95 - 2.32) for the
highest CRP category (> 3 mg/l), adjusted for all confounders (Table 2). Moreover, the
increased risk of developing COPD in subjects with the highest category of hsCRP was
confirmed in subanalyses, only including those participants with a spirometry investigation in
the context of the Rotterdam Study (Table E2 online data supplement). Per unit increase in the
logarithm of serum levels of CRP, a 20% increase in the risk of developing definite or
probable COPD, and a 30% increase in the risk of developing definite COPD during the
follow-up period was observed (Table 2).
Stratification by gender showed a comparable risk for men and women. Stratification
by smoking showed that the risk was most pronounced in the former smoker group (HR 2.2;
95% CI, 1.12 - 3.74). The risk was less pronounced in current smokers (HR 1.1; 95% CI, 0.67
- 2.08), and hsCRP serum levels were not a risk factor in subjects who had never smoked (HR
0.8; 95% CI, 0.28 - 2.67) (Table 3). The interaction term (smoking behaviour)*(hsCRP) was
significantly associated with incident COPD (p<0.001).
Page 10 of 39
The three tagging SNPs of the CRP gene were in Hardy Weinberg equilibrium (1184 C > T ,
χ2 = 0.319, P = 0.57; 2042 C > T, χ2 = 3.337, P = 0.68; 2911 C > G, χ2 = 3.844, P = 0.05).
Firstly, we investigated the association between the individual CRP SNPs and incident
COPD. Heterozygosity of 1184 T was associated with a 1.3 fold increased risk of COPD
(95% CI, 1.06 - 1.69). However, the association decreased and became not significant for the
homozygous 1184 TT genotype and might therefore be a spurious result (Table 4). No
associations were found in the subanalyses stratified by gender and smoking status (data not
Secondly, we studied the four CRP haplotypes. Haplotypes 2, 3 and 4 provided all
significantly higher CRP levels than haplotype 1 (18). Comparable to the individual SNPs,
there was an association between incident COPD and heterozygosity of haplotype 2 (based on
SNP 1184), but like in the analyses with the individual CRP SNPs, the association decreased
and became non-significant for the homozygous haplotype (Table 5).
Page 11 of 39
To our knowledge, this is the first study examining the association between serum hsCRP,
CRP gene polymorpisms, CRP gene haplotypes, and incident COPD, sampled in a population
without prior history of COPD. We demonstrated that increased serum levels of hsCRP were
associated with a 70% increase in the risk of developing COPD. This increased risk remained
significant after adjustment for several confounders, including cardiovascular co-morbidity,
and remained increased, although not significant, after introducing a 3-year latency period,
supporting the strength of the observed association. No strong evidence was found for an
association between variation in the CRP gene and incident COPD. Because we found that
hsCRP was not a predictive marker for COPD in subjects who had never smoked, we
conclude that baseline hsCRP levels cannot be considered as an independent risk factor for
Several cross sectional studies have shown that high levels of CRP are associated with
COPD (12, 29-31). Our results suggest that hsCRP levels raise early in the disease process,
before COPD is symptomatic. These results are supported by the fact that the introduction of
a 3-year latency period in our analyses hardly changed the hazard rates. The risks remained
significant after adjustment for several confounders, such as concomitant cardiovascular
disease, which are also associated with elevated levels of hsCRP. At least two mechanisms
can explain the observed association between increased hsCRP levels and incident COPD.
Firstly, the CRP levels could be raised because of subclinical COPD, as was suggested by
Aronson et al. (32). In their study, they demonstrated an inverse linear relationship between
CRP concentrations and FEV1. Secondly, high CRP levels could be associated with an
increased susceptibility to develop COPD. CRP is a classical acute-phase protein that binds to
ligands exposed in damaged tissue.
Page 12 of 39
This leads to the activation of the complement cascade, which results in a complement-
mediated exacerbation of tissue injury, as was previously shown for ischemic necrosis in heart
attacks and strokes (33). By this mechanism, subjects with a susceptible inflammatory profile,
could, when exposed to noxious gases or particles, have a higher risk to develop COPD.
Interestingly, the association between serum hsCRP and incident COPD appeared to
be highest in the former smokers. Previous studies demonstrated the existence of persistent
inflammation in COPD patients even after smoking cessation (34). In the group of persons
who were still smoking at the time of CRP sampling, the association between hsCRP and
incident COPD was less pronounced than in former smokers, possibly because current
smokers were relatively more healthy and had less (respiratory) complaints than former
smokers at the time of the interview and therefore did not stop smoking yet (i.e. the “healthy
smoker effect “ (35)).
Despite the growing scientific consensus regarding a prominent role of systemic
inflammation in COPD, little is known about the underlying mechanisms (8). Indication for a
primarily genetic origin (36, 37) comes from the fact that only 20% of the smokers seem to
develop the disease, suggesting the presence of COPD susceptibility genes. Silverman et al.
have published several reports regarding the importance of genetic variability in the
development of COPD (38). However, this is the first study of the influence of variation in the
CRP gene on the occurrence of COPD, although several studies have already given proof for
the heritability of serum CRP levels, even in COPD populations (17). Our results do not
support a role for genetic variability of the CRP gene as a cause for COPD, despite the fact
that haplotypes of the CRP gene influence hsCRP serum levels, since no association between
CRP gene haplotypes and incident COPD could be established, even in subanalyses stratified
by gender and smoking behaviour.
Page 13 of 39
A power calculation for our study showed that we were able to demonstrate relative risks for
COPD of at least 1.3 for the most common CRP SNPs 1184 and 2042, and 1.5 for the more
rare CRP SNP 2911 (39). Therefore, either there was indeed no association between the CRP
gene haplotypes and COPD, or the relative risks were of relatively small magnitude.
Other genetic and/or environmental determinants might contribute to the risk of
incident COPD associated with increased serum hsCRP levels, since several inflammatory
mediators (e.g. IL-6 or TNF-α) induce the formation of acute phase reactants such as CRP.
Similar to our findings, no association was found between CRP genotypes and arterial
thrombosis and coronary heart disease, despite the fact that CRP levels were a potent
predictor of future cardiovascular events in those same cohorts (18, 21). Moreover, a recent
publication by Tennent et al. showed no influence of transgenic expression of human CRP on
atherogenesis in a mouse atherosclerosis model, arguing against a causal role for CRP in
cardiovascular disease (40). Yet, the topic remains controversial, since another report from
Lange et al. described a positive association between variation in the CRP gene and the risk of
The findings of this study are important since several anti-inflammatory drugs already
exist and are successfully used in other medical specialities (33), supporting the possibility of
early intervention in COPD. The strengths of this study are the prospective, population-based
cohort design and the large number of study subjects (> 6000), with identical data collection
procedures for every participant. In this cohort, more than 600 cases of incident COPD were
identified over a total follow-up time of 15 years (1989-2004). All data collection during
follow-up of the cohort was done in a prospective manner, without knowledge of future
disease or research hypothesis, making selection and information bias unlikely.
Page 14 of 39
A potential limitation of our study is the single CRP measurement in time. However,
previous studies showed that the CRP level remains stable over time in normal individuals
and in patients with COPD (29). Also, the effect of confounding was probably limited, since
adjustment for all major confounders known to influence hsCRP levels in our analyses hardly
affected the risk estimates. Participants with CRP levels higher than 10 mg/l and subjects who
used systemic corticosteroids were excluded from the principal analysis in order to reduce the
effect of acute or chronic infections and other inflammatory processes, which can be more
often present in COPD patients. Still, when we performed additional analyses including these
participants, the results did not change materially (data not shown). In addition, we studied
CRP gene haplotype variants that affect CRP concentration levels and thus may reflect
lifetime exposure more accurately than the CRP serum concentration measured at a single
point in time (23).
In our study, we used a fixed value (0.7) of the FEV1/FVC ratio to define airflow
obstruction, rather than the fifth percentile of the predicted value (41). Literature has shown
that the lower limit of the normal range of the FEV1/FVC ratio decreases with age, meaning
that we would overestimate the number of patients with COPD if we would only use this
criterion (42, 43). To correct for this misclassification bias in our elderly cohort, we only
defined those subjects with a FEV1 < 80% as “definite” COPD cases, since Lindberg and al.
found that this cut-off seemed to be more reliable than the GOLD criteria in identifying
incident COPD among elderly subjects (44, 45). Also, a recent study showed that subjects
classified as “normal” using the lower limit of normal, but as “abnormal” using the fixed ratio
of < 0.7, were more likely to die and have COPD-related hospitalisations during further
follow-up than healthy subjects (46). This effect was even more pronounced among those
subjects with a FEV1 < 80%.
Page 15 of 39