Prevalence and correlates of inflammatory arthritis in Germany: data from the First National Health Survey
- First Online:
- Cite this article as:
- Schneider, S., Schmitt, G. & Richter, W. Rheumatol Int (2006) 27: 29. doi:10.1007/s00296-006-0153-0
- 127 Views
The aim of this paper is to generate data on the prevalence of inflammatory arthritis in different subgroups of the population and to identify correlates on the basis of population-based cross-sectional data: the “First National Health Survey of the Federal Republic of Germany”. This Survey investigated the prevalence of inflammatory arthritis, comorbidity and health-relevant behaviors on the basis of interviews with physicians and medical evaluations conducted in the period from October 1997 to March 1999. The study was based on a net sample comprising 6,461 subjects aged 18–79. Our data demonstrate an overall prevalence of 3.4% for inflammatory arthritis. The prevalence of inflammatory arthritis is significantly higher in women, the over-50, lower-income groups, and habitual smokers. Patients with inflammatory arthritis have a higher rate of numerous comorbidities such as osteoporosis, thyroid disease, chronic bronchial disease, hypertension, and elevated blood lipids versus healthy reference groups.
KeywordsInflammatory arthritisSurveyEpidemiologic determinantsCorrelatesPrevalence
In the vast majority of cases, inflammatory arthritis is a chronic autoimmune disease of unknown origin, probably multifactorial. The most common diseases in this category are rheumatoid arthritis and ankylosing spondylitis. In Germany rheumatoid arthritis is estimated to cost EUR 5,000 in direct hospital costs, not counting indirect costs of EUR 10,000 per patient per year [1–3]. The direct and indirect cost of ankylosing spondylitis is estimated at approximately 50% of the figure for rheumatoid arthritis . The expenses associated with hospitalization account for approximately half of the total direct cost, while sick leave accounts for approximately two-thirds of the indirect cost . The total cost of disease for inflammatory rheumatic conditions is 1.5–2 times that of other major chronic illnesses such as diabetes and chronic obstructive pulmonary disease .
Current epidemiological studies on the prevalence and correlates of inflammatory arthritis are available for USA, UK, France, Italy, Norway, and Finland [5–9]. For Germany, no representative data existed to date on prevalence and correlates of inflammatory arthritis. The Federal Statistics Office and Ministry of Health base hitherto their health reports on selective data from the German Rheumatology Research Center in Berlin, which compiles patient data from 100 rheumatology centers and offices. Given that the prevalence of inflammatory arthritis is likely to rise in the near future on the basis of demographic trends, this paucity of data in the European Union’s most highly populated member state is rather critical.
Epidemiological screening for correlates promises to be useful in view of the fact that environmental factors may compound the undisputed effect of genetic predisposition in helping to trigger the disease [5, 10]. Environmental factors can be broken down into hormonal, lifestyle-specific, and infectious disease-related determinants [7, 11]. Many authors emphasize the usefulness of epidemiological studies for the identification of potentially relevant environmental factors as a first step in identifying high-risk groups [10, 12].
The aim of this paper is therefore to generate up to date representative data on the prevalence of inflammatory arthritis in different subgroups of the population and to identify potential correlating factors on the basis of population-based cross-sectional data from the German National Health Survey.
Materials and methods
The National Health Survey is a representative epidemiological study of the population of the Federal Republic of Germany. This cross-sectional study was the first investigation of the prevalence of inflammatory arthritis, their correlates, and several other morbidities in post-reunification Germany. The Robert Koch Institute, Berlin, performed the Data acquisition, on behalf of the German ministry of health. The Orthopedic Department of Heidelberg University Hospital conducted the data analysis for the present study as part of an interdisciplinary research effort . The survey was conducted in the period from October 1997 to March 1999 and involved a total net sample size of 7,124 people aged 18–79 with their main place of residence in the Federal Republic of Germany, corresponding to a participation rate of 61% [14, 15]. The factor weighting according to age, gender, community size, and federal state employed in the NHS enables representative conclusions to be drawn for the adult population in Germany. After excluding incomplete datasets, a sample of 6,461 people was available for the analysis described below. The medical interviews and evaluations of these 6,461 subjects were carried out at 130 sites in 113 cities by four mobile inspection teams composed of physicians and medically qualified examiners [14, 16, 17]. In addition to completing a self-administered questionnaire containing items pertaining to medical factors and health-related behaviors, subjects underwent a standardized physician interview and medical evaluation during a period of several hours at the study site .
Documenting diseases and potential confounders
The outcome measure of inflammatory arthritis and other diseases were assessed by a physician’s interview. This included questions and information on lifetime prevalence of 44 clinical entities. Inflammatory arthritis was defined as being present in respondents who said they suffered from a medically diagnosed (e.g., family doctor diagnosed) inflammatory arthritis and had experienced symptoms of the disease during the four weeks prior to the date of the medical interview. To help circumscribe the definition for investigators “rheumatoid arthritis” and “ankylosing spondylitis” were given as specific examples of inflammatory rheumatic disease. Degenerative joint disease (osteoarthritis), dorsopathy, and pararheumatic conditions (osteoporosis, gout) were explicitly excluded from the “inflammatory arthritis” category and recorded separately elsewhere. This procedure is in compliance with the classification of “inflammatory arthritis” adopted by the German Rheumatic Disease League . Comorbidities were quantified by documenting the prevalence of relevant (e.g., family doctor diagnosed) concomitant illnesses, hypertension, and hyperlipidemia at the time of the physician’s interview. Subjects with inflammatory arthritis were also asked whether these comorbidities were present before the arthritis was first diagnosed.
The respondent’s height and weight were measured in the course of the medical examination using calibrated instruments, without shoes and outer clothing. The results were used to calculate body mass index on the basis of the interval limits used in studies by Uhlig et al.  and Symmons et al. . Current tobacco use was classified in accordance with WHO MONICA criteria according to the categories, smoker (daily tobacco use), occasional smoker (less than daily tobacco use), former smoker, and never smoked . The average quantity of pure alcohol consumed daily was calculated on the basis of six quantity (milliliters) and seven frequency categories. Respondents who reported consuming caffeinated coffee on a daily basis during the past year were contrasted with the remaining respondents. Daily consumption of black tea was analyzed the same way. Vegetarianism was defined as a completely or mainly meat-free diet.
In the occupation category, non-earners at the time of the interview were classified according to their most recent occupational position . Socioeconomic status was constructed as an additive index generated on the basis of income, occupational position, and highest educational qualification . Analysis of the socioeconomic factors “occupational position”, “educational qualifications”, and “socioeconomic status” is extensively documented and validated elsewhere . Each respondent’s infectious disease history (covering typhoid fever, paratyphoid fever, and childhood illnesses) was documented on the basis of information elicited from the subject.
A bivariate analysis was first conducted to determine percentage prevalence rates for individual risk groups. The chi square test was applied to investigate intergroup differences. The impact of hormonal, infectious disease, and lifestyle correlates was tested by logistic regression analysis on the basis of dummy coding of the dependent variables. Adjusting age and gender and calculating odds ratios kept possible confounders constant. To control the simultaneous effect of the statistically significant correlates and any spurious correlations, a final multiple logistic regression analysis was conducted with all hitherto non-significant variables excluded. This procedure corresponds to the statistical design used by Uhlig et al.  in another population-based arthritis study using identical age ranges. To avoid collinearity, the stated occupation was selected from the three status indicators for the regression analysis. All tests were two-tailed using the P ≤ 0.05 level of significance. All analyses were done using the statistics program SAS for Windows, Version 9.1.3 (SAS Institute Inc. Cary, NC, USA).
Environmental factors as correlates of inflammatory arthritis
Representative nationwide data for Germany on correlations between hormonal factors and inflammatory arthritis
Prevalence in %
Odds ratio (95% CI) adjusted for age and gendera
Odds ratio (95% CI) adjusted for all significant factorsb
Underweight (< 20)
Normal weight (BMI 20–< 25)
Overweight (25–< 30)
Obesity (≥ 30)
Representative nationwide data for Germany on correlations between lifestyle-specific factors and inflammatory arthritis
Prevalence in %
Odds ratio (95% CI) adjusted for age and gendera
Odds ratio (95% CI) adjusted for all significant factorsb
(0.89 ; 1.81)
(0.88 ; 1.84)
(0.65 ; 2.37)
(0.76 ; 2.80)
(1.05 ; 2.15)
(0.98 ; 2.05)
Level of alcohol consumption
Low ( < 5 g/day)
Moderate (5 - < 30 g/day)
(0.78 ; 1.51)
High ( ≥ 30 g/Tag)
(0.69 ; 2.11)
(0.81 ; 1.52)
(0.83 ; 1.64)
(0.41 ; 1.61)
(1.03 ; 2.23)
(1.09 ; 2.37)
Manual workers and farmers
Foremen and master artisans
(0.43 ; 0.96)
(0.42 ; 0.92)
Office workers and middle grade civil servants
(0.31 ; 0.68)
(0.29 ; 0.64)
Self-employed with 9 employees or fewer
(0.19 ; 1.15)
(0.16 ; 0.98)
White-collar workers and senior grade civil servants
(0.33 ; 0.89)
(0.29 ; 0.81)
Top-level office workers and civil servants, executive management
(0.07 ; 1.06)
(0.06 ; 0.88)
Students, trainees, unskilled workers
(0.48 ; 1.31)
(0.50 ; 1.40)
No school-leaving qualifications or Hauptschul diploma
Middle school, apprenticeship
(0.55 ; 1.03)
High school diploma (Abitur)
(0.36 ; 1.44)
(0.29 ; 0.86)
(0.63 ; 1.17)
(0.34 ; 0.84)
(0.72 ; 1.77)
(0.75 ; 2.00)
(0.64 ; 1.53)
A final logistic regression analysis (Tables 1, 2) was then used to test (with all significant correlates factored in) the extent to which these factors disappear if other variables are included and kept constant. Thus, the typical arthritis patient is 50 years old, female, a habitual smoker, and has (or had) a manual occupation. Conversely, the arthritis prevalence is significantly lower among younger male non-smoking office workers. The synopsis also shows that the known age gradient is preserved in the multiple logistic regression analysis if confounders are held constant. The disease prevalence in the over-50 in our study is around 5–7 times higher than in the reference group of 18–29 year-olds (1.0% vs. 5.5–6.3%; Table 1). The figures also confirm that women have a twofold higher prevalence of inflammatory arthritis compared with men (2.3 vs. 4.5%, Table 1). The odds ratios for all occupational groups are below 1.0 in relation to manual (skilled and unskilled) workers and farmers. Thus, the disease prevalence was lower for office workers, intermediate-grade civil servants and for senior grades (Table 2). Even after adjustment for occupational risk exposure, the prevalence of inflammatory arthritis in the former West German states is still much higher than in the territories of the former East Germany.
Infectious diseases as correlates of inflammatory arthritis
Representative nationwide data for Germany on correlations between infectious disease-related factors and inflammatory arthritis
Prevalence in %
Odds ratio (95% CI) adjusted for age and genderb
Test level/sigificance level
χ² = 17.3
P < 0.001
(0.88 ; 2.07)
χ² = 7.6
P = 0.006
(0.98 ; 1.78)
χ² = 3.3
P = 0.070
(0.84 ; 1.52)
χ² = 2.9
P = 0.090
χ² = 0.3
P = 0.565
χ² = 14.8
P < 0.001
χ² = 2.7
P = 0.099
χ² = 7.0
P = 0.008
Comorbidites of inflammatory arthritis
Validity and representativeness of the database
Our data demonstrate an overall prevalence of 3.4% for inflammatory arthritis. The figure is somewhat higher than the figures previously quoted for Germany, which put the cumulative prevalence of rheumatoid arthritis and ankylosing spondylitis at 2.0–3.0% [1, 2, 4, 21, 22]. There are a number of possible reasons for the higher prevalence figures in the National Health Survey. Harris for instance points out that prevailing demographic trends (the main one being a higher life expectancy) mean that current prevalence statistics are likely to be higher than those published in the past . Other reasons we can identify are the non-inclusion of, barely affected, children, and adolescents in the sample [7, 10], and the operationalization of inflammatory arthritis . A criticism that applies in this and other studies [24–26] is that it was not possible to conduct examinations and diagnostic procedures (for example using ACR classification criteria; ) neither with regard to inflammatory arthritis nor with regard to comorbidities (for example, bone densitometry to confirm a diagnosis of osteoporosis; ) or to validate the diagnosis e.g., by the call to the primary care provider or the specialist of the participants. The fact that the infectious disease history is based on retrospective self-reports is another limitation of this study.
The validity and representativeness of numerous other survey data were likewise reviewed and documented on the basis of intensive internal and external quality assurance: Half of the non-participants were prepared to provide basic health information and socioeconomic data for a nonrespondent analysis. According to the data provided, non-participants, and participants did not differ with regard to age, gender ratio, smoking or health, but non-participants were more likely to have lower educational qualifications . In Germany, there is a paucity of valid longitudinal studies on inflammatory rheumatic disease that the present survey cannot hope to remedy .
Cross-sectional clinical trials in patient cohorts for their part are associated with a risk of selection bias in that comparatively severe cases are more likely to be over represented in the study population [5, 8]. The pros and cons of cross-sectional arthritis studies like this one have been abundantly discussed [28, 29], and it is clear that the significant correlates of inflammatory arthritis studied here cannot be interpreted as causal in view of the cross-sectional study design. Interaction with the disease should be considered with regard to occupational stress factors and certain lifestyle aspects (including BMI, consumption, and behavioral patterns). For example, the diagnosis and resultant psychological stress might prompt patients to increase their tobacco consumption, although Norwegian data demonstrate that at least as many patients in that situation deliberately reduce their tobacco consumption for health reasons .
In agreement with our empirical findings, all relevant sources in the literature report an age-related increase in arthritis prevalence [7–9, 30]. Our data also reflect the phenomenon reported from the USA  and Greece  according to which the arthritis prevalence increases up to age 50, giving way to a more moderate rise thereafter (Table 1). The two-fold higher risk of disease in females as opposed to males is well documented in epidemiological studies [8, 30] and has been attributed to hormonal factors, although the specific underlying mechanisms are entirely unclear at this time [7, 10, 11]. Our age and gender-specific prevalences are almost identical to US figures  showing that women exhibit a sharply increasing arthritis prevalence to coincide with the typical menopausal age while prevalences in the male population stagnate in the over-50 age groups (Fig. 1). Symmons advances an explanation  based on the theory that the protective effects of oral contraceptives no longer apply after the menopause.
Several international studies also document subfertility as a significant factor in chronic inflammatory arthritis; childless women exhibited a higher morbidity risk in three out of eight studies [7, 10, 11]. Silman et al.  states in this connection that confounding social factors seem to play a fairly subordinate role here, as people living alone per se did not exhibit a significant difference in risk, even after adjustment for age. Corroborating Uhlig’s studies , our data likewise demonstrate a non-significant effect of family status and support the same conclusion (Table 2). Accordingly, childlessness is counted as a hormonal risk factor although the precise reasons are unknown [10, 11]. The German National Health Survey generated data on this aspect only by asking whether there were children living in the household. There is a risk of misclassification bias: adoptions may have diluted the effect identified. On the other hand, a number of women may be recorded as childless who have a child living elsewhere or who have had a termination of pregnancy, a miscarriage, a stillbirth or a child died in infancy. Given this methodological imprecision, the true effects may in fact have been underrated, resulting in more conservative outcomes.
Another factor is overweight as indicated by the BMI. As in other studies, the odds ratios in Table 1 demonstrate a U-shaped correlation between obesity and inflammatory arthritis . On the one hand, the slightly higher prevalences of underweight subjects (OR 1.1) may reflect disease-related weight loss . On the other hand, there is an increase in prevalence for BMIs of 25 and over. Silman et al.  postulates a confounding effect where increased estrogen production promotes both overweight and inflammatory arthritis. Unlike other authors , we adjusted the odds ratios for gender and age, which receded below the level of significance as a result (Table 2). Uhlig et al.  and Heliovaara  followed the same procedure and achieved the same result in various population-based cohorts [8, 9, 32].
Dietary and consumer habits have been debated in the literature as representing systemic lifestyle-specific factors. The database is fairly conclusive as regards the harmful effects of habitual tobacco consumption [9, 11, 32]. Nine out of the 13 studies published to date demonstrate a risk-elevating effect . Unlike those of other authors [8, 32, 33], our results are based on adjustment of other lifestyle variables. This rules out a spurious correlation in which smoking acts as a proxy for an unhealthy lifestyle overall. A biologically founded medical reason for this negative effect remains to be identified [8, 11, 28].
The conclusion that consumption of alcohol, coffee, or tea, or other dietary habits, has no effect on the development of inflammatory arthritis is supported by our data and the studies authored by Reckner et al.  and Heliovaara et al. . In response to contradictory results from Finnish data, Heliovaara et al.  pointed out that coffee consumption may stand as a partial proxy for an unhealthy lifestyle overall. In view of the cross-sectional nature of the study design, it is possible that effects in the other studies and our own are diluted by the fact that patients may adjust their diet after being diagnosed [7, 8].
From a bivariate point of view, some studies indicate a negative correlation between occupational, educational, and socioeconomic indicators on the one hand and arthritis prevalence on the other. This effect is explained by poorer medical care, unhealthier working conditions and an unhealthier lifestyle among lower socioeconomic groups [5, 8, 23, 32, 34, 35].
Infectious disease-related factors and comorbidities
Numerous disease-causing organisms have been linked with a risk of developing inflammatory arthritis later in life [5, 11]. A statistical relationship, but no causal connection, has been demonstrated for some childhood diseases (Table 3). The non-significant odds ratios obtained when age is held constant are due to lower infectious disease prevalences and concomitantly, but independently, lower inflammatory arthritis prevalences in younger subjects.
More research has also been called for with regard to the relevance of comorbidities in subjects with inflammatory rheumatic disease [11, 36]. Some authors have postulated thyroid disease, gout, and other metabolic disorders such as hypertension, hyperlipidemia, and diabetes as comorbidities of inflammatory arthritis [5, 11, 32, 37]. As cutoff date-based data of this kind cannot be interpreted as indicating coetiology, additional data were generated to determine which comorbidities were already present before the inflammatory arthritis was first diagnosed. Additional analyses not presented here showed that the prevalence of gout, hyperlipidemia, and hypertension was significantly higher (i.e., already present at the time of diagnosis) in the population of subjects who went on to develop inflammatory arthritis (P < 0.05, age-adjusted).
To sum up: the German National Health Survey provides the first representative cross-sectional data on inflammatory rheumatic disease and its correlates for post-reunification Germany. In this country, a period of 1.6 years elapses on average from disease onset to first contact with a rheumatologist or rheumatology center. The corresponding figure for ankylosing spondylitis is as high as 4.9 years . The data presented here may make basic epidemiologic data available to specialists and generalists and help to show physicians the comorbidities and correlates that should be considered in the diagnosis of arthritic subjects.
Special thanks are due to Dr. Heribert Stolzenberg, Robert Koch Institut, Berlin for the providing the data sets. We also wish to thank Saskia Tönges for her assistance in drafting and formatting the text. None of the authors has a conflict of interest. A grant of the independent research fund of the Department of Orthopaedic Surgery, University of Heidelberg, Germany supported this publication.