Undiagnosed Diabetes Mellitus in Community-Acquired Pneumonia: A Prospective Cohort Study

Abstract

Background

Diabetes mellitus is an important risk factor for community-acquired pneumonia, whereas the prevalence of undiagnosed diabetes mellitus and prediabetes in patients with community-acquired pneumonia is largely unknown. We aimed to determine the prevalence of prediabetes, undiagnosed diabetes mellitus, and risk factors associated with undiagnosed diabetes mellitus in a large European community-acquired pneumonia cohort.

Methods

This was a multicenter prospective cohort study of hospitals and private practices in Germany and Austria encompassing 1961 adults with community-acquired pneumonia included in the German Community-Acquired Pneumonia Competence Network (CAPNETZ) study between 2007 and 2014. The prevalence of undiagnosed diabetes mellitus and prediabetes was estimated based on hemoglobin A1c measurements. Logistic regression was used to assess risk factors for undiagnosed diabetes mellitus.

Results

Fifteen percent of patients had known diabetes mellitus. Among patients without known diabetes mellitus, 5.0% had undiagnosed diabetes mellitus and 37.5% had prediabetes. Male sex (odds ratio [OR], 2.45 [95% confidence interval {CI}, 1.35–4.45]), body mass index ≥25 kg/m² (OR, 2.64 [95% CI, 1.48–4.72]), and hyperglycemia at admission (6–11 mM: OR, 2.93 [95% CI, 1.54–5.60] and ≥11 mM: OR, 44.76 [95% CI, 17.58–113.98]) were associated with undiagnosed diabetes mellitus. Patients with undiagnosed diabetes mellitus had a higher 180-day mortality rate compared to patients without diabetes mellitus (12.1% vs 3.8%, respectively; P = .001).

Conclusions

Undiagnosed diabetes mellitus was prevalent among community-acquired pneumonia. Male sex, overweight, and hyperglycemia at admission were associated with undiagnosed diabetes mellitus. The long-term mortality among patients with undiagnosed diabetes mellitus was high compared to patients without diabetes mellitus.

Community-acquired pneumonia (CAP) is a leading cause of hospitalization and mortality worldwide [1, 2]. Despite improvements in the treatment of CAP, the mortality has barely changed during the last 50 years [3–5], which holds true for both short- and long-term mortality [3, 6–8]. The high long-term mortality of CAP has been related to comorbidity [6, 9] and, therefore, studies focusing on CAP and the impact of comorbidities are warranted.

It has been estimated that 415 million people worldwide have diabetes mellitus (DM), and the International Diabetes Federation has estimated that every second patient with DM remains undiagnosed [10]. Usually, DM is associated with conditions such as cardiovascular disease, neuropathy, and kidney disease, but DM is also associated with infections and a risk factor for fatal outcome in CAP [11].

With DM being a risk factor for pneumonia [12], CAP could be an opportunity to test for undiagnosed DM. However, little is known about the current burden of pre-DM and unrecognized DM in CAP patients. Early DM diagnosis followed by antidiabetic treatment and lifestyle interventions may reduce acute and long-term complications [13, 14] and be a cost-effective intervention [15]. Our main objective was therefore to assess the prevalence of pre-DM, undiagnosed DM, and risk factors associated with undiagnosed DM in a large European CAP cohort (the German Community-Acquired Pneumonia Competence Network [CAPNETZ]; www.capnetz.de).

METHODS

Study Design, Setting, Population, and Data Collection

Patients were prospectively recruited from 17 clinical centers in Germany (n = 16) and Austria (n = 1) into the CAPNETZ cohort [16]. Inclusion criteria were age ≥18 years, a pulmonary infiltrate diagnosed by chest radiography, and at least 1 of the following criteria; history of fever; cough; or purulent sputum production or focal chest signs on auscultation. Exclusion criteria were hospital admission within the past 28 days; immunosuppression, defined as chemotherapy or neutropenia <1000 μL during the past 28 days; treatment ≥20 mg corticosteroids daily for ≥14 days; human immunodeficiency virus infection; immunosuppressive therapy after organ or bone marrow transplantation; and active tuberculosis.

All clinical and score parameters were evaluated at first contact and stored in an electronic database. Serum glucose levels were measured on admission. CURB-65 was used as a severity score [17]. CURB-65 score is based on 5 factors: Confusion, urea >7 mmol/L, respiratory rate ≥30 breaths/minute, systolic blood pressure <90 mmHg or diastolic blood pressure ≤60 mmHg and age ≥65 years [17]. Patients were followed for 180 days. All patients included in the CAPNETZ cohort in 2007–2014 and with an available whole blood sample were included.

Variables

Measurement of Hemoglobin A1c

Hemoglobin A1c (HbA1c) was analyzed on whole blood using high-performance liquid chromatography (TOSOH G7). Samples were stored at –80°C prior to analysis. TOSOH G7 validates the quality of each sample by calculating the theoretical plates [18]. If the theoretical plate was <250, the sample quality was poor and it was considered as a nonvalid measurement. The analyses were performed at Steno Diabetes Center A/S, Copenhagen, Denmark. HbA1c is reported as mmol/mol, according to the International Federation of Clinical Chemistry.

Definition of Known DM, Undiagnosed DM, Pre-DM, Overweight, and Obesity

Known DM was defined on the basis of self-reporting and/or use of antidiabetic medicine. In patients without known DM, undiagnosed DM was defined as admission HbA1c ≥48 mmol/mol (≥6.5%) [19]. Pre-DM was defined as admission HbA1c 39–47 mmol/mol (5.7%–6.4%) according to the American Diabetes Association [19]. Patients with HbA1c 42–47 mmol/mol (6.0%–6.4%) where classified as pre-DM with very high risk of progression to DM [19]. As in previous studies [11, 20], we defined categories for admission blood glucose concentration of (<108.00 mg/dL), (108-197.9 mg/dL), and ≥198 mg/dL. According to the World Health Organization, we defined overweight and obesity as body mass index (BMI) ≥25 kg/m² and BMI ≥30 kg/m², respectively.

Outcome Measures, Exposures, and Potential Source of Bias

The primary outcome was the prevalence of undiagnosed DM, pre-DM, and risk factors associated with undiagnosed DM. Secondary outcome was 28-day mortality (short-term) and 180-day mortality (long-term). Age, sex, comorbidities present at admission, pneumonia severity according to CURB-65 [17], admittance blood glucose, and BMI were considered as primary predictors. CAP is associated with age and therefore we calculated an age-standardized prevalence of undiagnosed DM and pre-DM.

Statistical Analyses

Data were analyzed using SAS Enterprise Guide version 7.1 and Sigmaplot version 13 software. Data were described as count (%) for categorical variables and either mean (standard deviation) or median (interquartile range) for continuous variables as appropriate. The Wilcoxon rank-sum test, χ², Fisher exact test, and log-rank test were used to test for differences between participants with and without undiagnosed DM as appropriate. When comparing patients with known DM, undiagnosed DM, and without DM, P values were adjusted for multiple comparisons using the Bonferroni correction. We calculated an age-standardized prevalence of undiagnosed DM and pre-DM based on the age distribution in Germany in 2011 (http://ec.europa.eu/eurostat/web/main/home). We calculated the prevalence of undiagnosed DM in our age groups and multiplied them to the number of persons in each age group of the standard population, yielding the expected prevalence of undiagnosed DM in the standard population. We added up the number of expected prevalence of undiagnosed DM from each age group and divided this total number of undiagnosed DM by the total standard population, resulting in the age-standardized prevalence of undiagnosed DM [21].

Risk factors associated with unknown DM were identified using logistic regression models excluding patients with known DM. Variables associated with undiagnosed DM in the univariate analysis at a significance level of P < .10 were included in the multivariable analysis.

We calculated the number needed to test (NNT) to detect 1 patient with unknown DM: NNT = (patients at risk of undiagnosed DM / patients with undiagnosed DM in the different categories of admission blood glucose).

We used the receiver operating characteristic (ROC) to assess cutoffs for predictive associations between admission blood glucose and undiagnosed DM. HbA1c was used as reference for undiagnosed DM. Youden J statistic/index, [J = sensitivity + (specificity – 1)] was calculated to select the cutoff.

All statistical comparisons were 2-sided and carried out at the .05 significance level.

Ethical Considerations

Written informed consent was obtained from patients before study inclusion. The study was approved by the local ethical committees of participating centers.

RESULTS

In total, 2380 samples were available for analysis. Four hundred nineteen samples had a nonvalid measurement (17.6%); thus, data from 1961 patients were included (Figure 1).

Patients With a Nonvalid Measurement

The patients with a nonvalid HbA1c measurement did not differ compared to patients with a valid HbA1c measurement in terms of age, sex, current smoking, nursing home residency, BMI, CURB-65 score, and comorbidities, except for DM and pulmonary comorbidities other than chronic obstructive pulmonary disease (data not shown). Fewer patients with a nonvalid HbA1c measurement had known DM compared to the patients with a valid HbA1c measurement (9.6% vs 15.3%; P = .002). Four and a half percent had a blood glucose ≥11.0 mM compared to 8.0% in the patients with a valid sample (P = .03).

Characteristics of Patients With Undiagnosed Diabetes Mellitus and Without Diabetes Mellitus

Overall, 15.3% (300/1961) of patients had known DM and 74% (1450/1961) were inpatients. Among patients without a history of DM, 5.0% (95% confidence interval [CI], 4.0%–6.2%; 83/1661) had undiagnosed DM and 37.5% (95% CI, 35.2%–39.9%; 623/1661) had pre-DM. Compared to patients without DM, those with undiagnosed DM were older (71 years vs 58 years; P < .001), more likely to be a nursing home resident (3.6% vs 0.9%; P = .049), male (75.9% vs 53.4%; P < .001), overweight, (50.0% vs 30.6%; P < .001), or obese (29.3% vs 19.7%; P = .03), and had more comorbidities in terms of cardiac and kidney disease (Table 1). More patients with undiagnosed DM (19.8% vs 6.3%; P < .0001) had signs of moderate to severe pneumonia (CURB-65 score ≥3) and elevated admission blood glucose ≥11.0 mM (23.5% vs 1.3%; P < .001) compared to patients without DM.

Mortality Among Patients With Community-Acquired Pneumonia According to Diabetes Status

Compared to patients without DM, those with undiagnosed DM had a higher 180-day mortality (12.1% vs 3.8%; P = .001) whereas there were no differences between patients with undiagnosed DM and known DM (12.1% vs 5.3%; P = .11) (Figure 2). Among patients with undiagnosed DM, there were no differences between survivors and nonsurvivors, although the impression was that nonsurvivors were older (75.0 years vs 70.1 years; P = .08) and had higher median HbA1c (55 vs 50.0 mmol/mol; P = .25). There were no differences in the 28-day mortality between patients with undiagnosed DM, known DM, and without DM (1.2%, 2.7%, and 1.9% respectively; P = .61).

Prevalence of Undiagnosed Diabetes Mellitus and Prediabetes Stratified by Age Group

Undiagnosed DM was found in all age groups, except those ≤30 years. The prevalence of undiagnosed DM increased by age, with 1.2% (2/168) among 30–39 years and 9.1% (15/165) among patients ≥80 years (Table 2). The age-standardized prevalence of undiagnosed DM was 3.9%. Among patients without known DM, 37.5% (623/1661) had pre-DM and the age-standardized prevalence was 31.7%. Among patients with pre-DM, 40.1% (270/623) had HbA1c levels classified as “very high risk of progression to DM.” The prevalence of patients with very high risk of progression to DM increased by age, from 2.7% (4/149) among those aged 18–29 years to 29.1% (48/165) among patients ≥80 years (Table 2). A similar age distribution was seen in patients with pre-DM.

Risk Factors of Undiagnosed Diabetes Mellitus

In a multivariable logistic regression analysis (Table 3), male sex (odds ratio [OR], 2.45 [95%, CI, 1.35–4.45]) and BMI ≥25 kg/m² (OR, 2.64 [95% CI, 1.48–4.72]) were associated with undiagnosed DM. Hyperglycemia was associated with undiagnosed DM, and the OR for undiagnosed DM among patients with blood glucose ≥11 mM was 44.76 (95% CI, 17.58–113.98). The severity of CAP was associated with undiagnosed DM (OR, 1.99 [95% CI, 1.06–3.77] and 2.16 [95% CI, .96–4.83]) for a CURB-65 score of 2 and ≥3, respectively.

Number Needed to Test With Hemoglobin A1c to Diagnose 1 New Case of Diabetes Mellitus Stratified by Admission Blood Glucose

The NNT with HbA1c in our blood glucose categories ranged from 65 to 2 and the overall NNT was 20 (Table 4). Given the strong association between undiagnosed DM and blood glucose, we evaluated whether an admission blood glucose measurement could be indicative for undiagnosed DM. The ROC curve yielded an area under the curve of 0.79 (95% CI, .73–.85). The Youden J index suggested a blood glucose value of 7.4 mM as the cutoff for “diagnosing” DM with a sensitivity of 63% and a specificity of 83%. In this population, these numbers reflect a negative predictive value of 97.7%.

DISCUSSION

Undiagnosed DM and pre-DM was prevalent in patients with CAP. Admission hyperglycemia was strongly associated with undiagnosed DM. Known DM risk factors such as male sex and obesity were likewise associated with undiagnosed DM. Using HbA1c to test for undiagnosed DM in CAP patients yielded a low NNT.

The proportion of undiagnosed DM and pre-DM in the CAPNETZ cohort was 5.0% and 37.5%, respectively. In comparison, the only 2 other studies investigating the proportion of undiagnosed DM among CAP patients found 2% and 9%, respectively [22, 23]. Both studies were prospective, but generalizability was limited due to size and single-center design. In Spain, Falguera et al found a prevalence of undiagnosed DM of 2% (95% CI, 1.2%–3.3%) compared to the CAPNETZ cohort. They defined DM based on a fasting glucose of ≥7.0 mM or oral glucose tolerance test (OGTT) of ≥11.1 mM [22]. Falguera et al left it to the discretion of the attending physician to test for DM [22]. This may lead to a lower detecting rate of undiagnosed DM. In Finland, Salonen et al found a prevalence of undiagnosed DM of 9% (95% CI, 4.5%–13.2%) using an HbA1c of ≥48 mmol/mol [23]. The cohort comprised only patients with mild to moderate CAP. Due to small size (n = 153) and consequently wide 95% CIs, their estimate may not differ from the 5% found in the present study.

Due to the strong association between DM and risk of pneumonia [12], we expected that the prevalence of undiagnosed DM would exceed that of the general population. The prevalence of undiagnosed DM has been assessed in the German population (not CAP) [24–26], using various tests for DM (fasting glucose ≥7.0 mmol/L, OGTT ≥11.1 mM, nonfasting glucose ≥11.1 mM, or HbA1c ≥48 mmol/mol [≥6.5%]). Sex- and age-adjusted prevalence of undiagnosed DM ranged from 0.9% to 7.1% [24–26]. We determined the age-standardized prevalence of undiagnosed DM and, although the results cannot be compared directly to the aforementioned studies, the age-standardized prevalence of undiagnosed DM did not exceed that of the general population. We could speculate that the duration of DM and/or dysregulation of DM might influence the risk of acquiring pneumonia [27, 28]. Because long-term complications to DM develop over 5–20 years after disease onset [29] it may be that the effect of DM on the immune system also evolves over time. In a population-based study, Kornum et al found that patients with a DM duration ≥10 vs ≤10 years had increased risk of pneumonia-related hospitalization [30]. Patients with undiagnosed DM may have had DM for a short time period, as they have not presented with DM symptoms leading to the diagnosis. This may partly explain why we did not see a higher proportion of patients with undiagnosed DM. Also, the awareness of risk factors for DM among doctors and patients is high; hence, most DM may be detected early, leaving few patients undiagnosed.

Patients with undiagnosed DM had higher long-term mortality than patients without DM, whereas the short-term mortality did not differ. Whether the patients with undiagnosed DM have been diagnosed with DM and/or received antidiabetic treatment during follow-up is unknown. However, the patients with undiagnosed DM were older and more likely to be nursing home residents than patients without DM. This may have reduced the physicians’ inclination for performing DM diagnostic prior to the current admission. We can only speculate that undiagnosed DM and lack of antidiabetic treatment may be the cause of the higher mortality. It is noteworthy that the difference is only present at long-term follow-up, where comorbidities are known to contribute significantly to mortality [3, 9]. Detection and treatment of undiagnosed comorbidities may offer possibilities to reduce the high long-term mortality following CAP. Moreover, almost one-third of the patients with undiagnosed DM were current smokers and a lifestyle intervention following a DM diagnosis would bring an opportunity to promote smoking cessation.

To our knowledge, no other studies have reported the prevalence of pre-DM among patients with CAP. Recognition of pre-DM is important because individuals with pre-DM have a 5- to 10-fold higher annual risk of developing DM compared to normoglycemic individuals [31], and up to 70% progress to DM during their lifetime [32]. We found 37.5% with pre-DM and of these, 40% could be classified as “pre-DM with very high risk of progression to DM.” As with undiagnosed DM, the prevalence of pre-DM increased according to age. Most importantly, 10% of younger patients (≤30 years of age) had pre-DM. These patients could be of particular interest, as lifestyle interventions could have an impact on reducing their risk of DM [32, 33].

Known DM risk factors (ie, male sex and obesity) were associated with undiagnosed DM. Recent studies have shown that obesity is a risk factor for infection, including lower respiratory tract infection [34, 35]. The majority of those with undiagnosed DM were overweight (50%) or obese (29%). Given the association between overweight, risk of infection, and undiagnosed DM, it seems reasonable to pay special attention to these patients.

A blood glucose value of ≥11.1 mM and symptoms of hyperglycemia (eg, thirst, polyuria, weight loss, and blurry vision) are diagnostic of DM [36]. However, in patients with CAP, hyperglycemia may be due to physiological stress and hence transient, and DM should not be assessed with ongoing infection [37]. HbA1c is unlikely to be affected by acute infection [38] and we did find that blood glucose values ≥11 mM were strongly associated with undiagnosed DM. Also, blood glucose values ranging 6–11 were associated with undiagnosed DM. The association between hyperglycemia and DM, despite physiological stress, has also been demonstrated by MacIntyre et al among CAP patients [20]. The hazard ratio was 11 for developing DM within 5 years for a random admission blood glucose ≥11.1 mM [20].

We consider our NNT of 20 to be reasonably low and recommend testing for DM using HbA1c in adult patients with CAP. Our findings support the findings by MacIntyre et al. They reported, among patients with a blood glucose ≥6 mM and ≥11 mM, an NNT of 8 and 3, respectively, compared with 12 and 2 in our cohort [20]. From our ROC analysis, the blood glucose cutoff was 7.4 mM for detecting DM, yielding a high population-specific negative predictive value of 98%. In this CAP population, admission blood glucose levels <7.4 mM are likely to reject DM.

Strengths

Our patients were included prospectively from 17 centers. A standardized reporting system and database were used, ensuring a systematic recording of comorbidities. Both private practices and tertiary referral centers cooperated in patient recruitment. Thus, we believe the cohort to be representative of patients with CAP in developed countries. To our knowledge, this is the first study to evaluate the proportion and age-standardized prevalence of pre-DM and risk factors associated with undiagnosed DM in a large multicenter CAP cohort.

Limitations

HbA1c can be affected by conditions affecting red cell turnover (eg, hemolysis, major blood loss, or blood transfusions). We were not able to control for such conditions. We find it unlikely that our cohort should suffer from these conditions, as patients hospitalized within the 28 days prior to CAP diagnosis and patients with immunodeficiency were excluded. Although acute illness has little impact on HbA1c [38], we recommend further testing for DM after the pneumonia has resolved to confirm the diagnosis of DM. The agreement between HbA1c and fasting glucose or OGTT is imperfect, as is the concordance between fasting glucose and OGTT [19]. Data from the National Health and Nutrition Examination Survey showed that HbA1c detected one-third fewer cases of DM than fasting glucose or OGTT [39]. Hence, testing with HbA1c may give conservative estimates compared to glucose-based tests. To compare to the general population, we calculated the age-standardized prevalence of undiagnosed DM. We were not able to determine whether the prevalence of undiagnosed DM is unique to CAP or if it is a general trait in patients with infections. A total of 17.6% of the HbA1c measurements were nonvalid and were excluded from our analyses. We were not able to determine whether this was due to patient-related factors and/or to storage and sample handling, although HbA1c is proven to be stable when measured in stored whole blood samples [40]. Overall, patients with a nonvalid measurement had lower prevalence of known DM, which may lead to an underestimate of the prevalence of undiagnosed DM.

In conclusion, undiagnosed DM and pre-DM are prevalent in CAP. Despite ongoing infection, admission hyperglycemia was strongly associated with undiagnosed DM. One of every 10 patients aged ≤30 years had pre-DM. These patients could be of particular interest as lifestyle interventions could have a great impact on their risk of developing DM. The NNT for undiagnosed DM with HbA1c was low. The long-term mortality among patients with undiagnosed DM was high compared to patients without DM. We encourage clinicians to consider CAP as an opportunity to test for DM/pre-DM and refer for further counseling and treatment.

Notes

Author contributions. A. V. J. had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: A. V. J., D. F.-J., G. B. E., S. B. A., P. T. P., T. B., M. W., G. R., P. R. Acquisition, analysis, or interpretation of data: A. V. J., D. F.-J., G. B. E., P. T. P., T. B., G. R., P. R. Drafting of the manuscript: A. V. J., D. F.-J., T. B., G. R., P. R. Critical revision of the manuscript for important intellectual content: D. F.-J., G. B. E., S. B. A., P. T. P., T. B., M. W., G. R., and P. R. Statistical analysis: A. V. J. Obtained funding: A. V. J., G. R., M. W., P. R. Administrative, technical, or material support: G. B. E., P. T. P., S. B. A. Supervision: D. F.-J., T. B., G. R., P. R.

Acknowledgments. We thank the doctors who saw and identified patients with CAP for their work dedicated to CAPNETZ; the CAPNETZ study team involved in patient recruitment and sample and data handling; and the patients included in the CAPNETZ study. Further we thank Professor Lise Tarnow (Nordsjællands Hospital) and Professor Jørgen Rungby (Bispebjerg Hospital) for commenting on the manuscript.

Disclaimer. The funding parties did not have any influence on the study design, execution of the study, or interpretation of the results.

Financial support. This work was supported by the Christenson-Cesons family Foundation; Fru Olga Bryde Nielsens Foundation; Kaptajnløjtnant Harald Jensens og Hustrus Foundation; Department of Pulmonary and Infectious Diseases, Nordsjællands Hospital-Hillerød; the research council at Nordsjællands Hospital; and Bundesministerium für Bildung und Forschung (grant number FKZ 01ZX1304B) (CAPSyS). CAPNETZ was founded by a Bundesministerium für Bildung und Forschung grant (number 01KI07145 2001–2011).

Potential conflicts of interest. T. B. reports personal fees from GSK, Bristol-Myers Squibb (BMS), AbbVie, and Gilead as an advisory board member; personal fees from BMS and GSK as payment for lectures including service on speakers’ bureaus; personal fees from BMS and GSK as payment for development of educational presentations; and nonfinancial support from BMS and from Gilead for travel/accommodation/meeting expenses. G. R. reports personal fees from Pfizer, Boehringer Ingelheim, Solvay, GSK, Essex Pharma, MSD, and Novartis for lectures including service on speakers’ bureaus; and personal fees from GSK for travel/accommodations/meeting expenses. P. R. reports personal fees from MSD, AbbVie, and CSL Behring as an invited speaker; personal fees from Statens Serum Institut as a data and safety monitoring board member; and nonfinancial research collaboration with Astellas. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Additional members of the CAPNETZ Study Group M. Dreher, C. Cornelissen (Aachen); W. Knüppel (Bad Arolsen); D. Stolz (Basel); N. Suttorp, P. Creutz (Berlin, Charité); T. Bauer, T. Sabha (Berlin); W. Pankow, A. Lies, D. Thiemig (Berlin-Neukölln); B. Hauptmeier, S. Ewig, D. Wehde (Bochum); M. Prediger, S. Schmager (Cottbus); G. Höffken, M. Kolditz, B. Schulte-Hubbert, S. Langner (Dresden), T. Welte, G. Barten, M. Abrahamczik, J. Naim, W. Kröner, T. Illig, N. Klopp (Hannover); C. Kroegel, M. Pletz, J. Happe, J. Frosinski, J. Winning, A. Moeser (Jena); K. Dalhoff, K. Dageförde, K. Franzen, F. Hyzy, H. Schmieg, P. Parschke, P. Thiemann, J. Ahrens, T. Hardel (Lübeck); J. Drijkoningen (Maastricht); H. Buschmann, R. Kröning (Paderborn); H. Schütte (Potsdam), T. Schaberg, I. Hering (Rotenburg/Wümme); C. Kropf-Sanchen (Ulm); T. Illmann, M. Wallner (Ulm); O. Burghuber, G. Rainer (Vienna); and all study nurses.

References