Two national health examination surveys were conducted in Germany in 1998 and 2008–2011. Both used a nationwide two-stage clustered sample design with selection of study points based on community type and federal state and subsequent sampling of persons aged 18–79 years stratified by sex and age group from the local population registers [17, 18]. The German Health Interview and Examination Survey for Adults (DEGS1) 2008–2011 comprised a total of n = 7,115 persons examined at one of 180 study points, of whom n = 4,192 were first-time participants (response: 42 %) and n = 2,923 (response: 62 %) were former participants of the German National Health Interview and Examination Survey 1998 (GNHIES98). The net sample of the GNHIES98 consisted of n = 7,124 persons (response: 61 %) from 120 study points. The study was approved by the ethical committee of Charité University Medicine, Berlin, and by the Federal Commissioner for Data Protection and Freedom of Information. Informed written consent and assent were obtained from all participants.
Measurement and survey methods
BP was measured according to a standardized protocol which was almost identical in both surveys, except that the standard mercury sphygmomanometer (Erkameter 3000, Bad Tölz, Germany) employed in the GNHIES98 was replaced by an automated oscillometric Datascope Accutorr Plus device (Datascope Accutorr Plus, Mahwah, NJ, USA) together with a new set of manufacturer-provided cuffs and adapted cuff-selection-rules. The participants sat quiet and upright on a height-adjustable chair with their back supported, the right forearm was resting on a table at heart level, elbow slightly bent, legs uncrossed and feet firmly on the floor. The correct cuff size was determined with the upper arm circumference (AC) measured half way between the acromion and the olecranon. Both surveys used three cuff sizes. The cuff bladder dimensions (width × length) in DEGS1 were: 10.5 × 23.9 cm for ACs of 21–27.9 cm, 13.5 × 30.7 cm for ACs of 28–35.9 cm and 17 × 38.6 cm for ACs of 36–46 cm. The corresponding sizes in GNHIES98 were: 8 × 20 cm for ACs < 20 cm, 12 × 28 cm for ACs of 20–40 cm and 14 × 40 cm for ACs > 40 cm. The correct position of the cuff above the brachial artery was ensured with a mark on the cuff. Three blood pressure measurements were taken at 3-min intervals, following an initial 5-min resting period (GNHIES98: 3 min) after a non-strenuous part of the examination.
The participants were asked to bring along their medication of the previous 7 days and antihypertensive medication use was defined according to the WHO Anatomic Therapeutic Chemical Classification System (ATC): antihypertensive drugs (C02), diuretics (C03), beta-blockers (C07), calcium channel blockers (C08) and ACE inhibitors (C09). Hypertension was defined as: SBP ≥140 mmHg or DBP ≥90 mmHg or treatment with ATC-coded antihypertensive medication. However, the antihypertensive medication was only used for defining hypertension if the participants reported having hypertension since the indication for taking these drugs may be other than hypertension.
Body height and weight measurements also followed standardized procedures with the participants dressed only in underwear without shoes. Body height was measured with a portable stadiometer (Holtain Ltd., UK, precision: 0.1 cm) and body weight with a calibrated electronic scale (SECA, column scale 930, precision: 0.1 kg). BMI was calculated as weight in kilogram (kg) divided by height in meter squared (m2) and BMI was used to define non-overweight (BMI < 25 kg/m2), overweight (BMI ≥ 25 to <30 kg/m2) and obesity (BMI ≥ 30 kg/m2).
Information on lifestyle and socio-demographic variables were obtained with a self-administered questionnaire. Social status was determined using an index with information on school education and vocational training, occupational status and net household income (weighted by household needs) permitting classification into low, middle and high status groups . Alcohol consumption was calculated in gram/day (g/d) on basis of questions about consumption frequency and amount of beer, light beer, alcohol-free beer, wine and liquor. Alcohol in g/day was then divided into three classes according to German guidelines on tolerable upper intake levels: non-drinker (0 g/d), light drinker (men: >0 to 20 g/d, women: >0 to 10 g/d) and heavy drinker (men: >20 g/d, women: >10 g/d) . Smoking status was assessed by smoking frequency (daily, occasionally, no longer, never) and amount of cigarettes smoked per day and was categorized into current daily smoker (≥1 cigarette per day) or non-smoker (including occasional smoker and ex-smoker). Sports activity was asked with “How often do you exercise?” and the response items were: “no sports activity”, “<1 h/week”, “1-2 h/ week”, “2-4 h/week” and “>4 h/week”. This information was subsumed into three groups: no sports activity, sports activity <2 h/week and sports activity >2 h/week.
Analyses were performed in 18–79 year old GNHIES98 and DEGS1 participants. Exclusion criteria were missing information on BP (GNHIES98: 0.2 %, DEGS1: 0.3 %), antihypertensive medication use (GNHIES98: 0.4 %, DEGS1: 0.3 %) or BMI (GNHIES98: 0.7 %, DEGS1: 0.7 %). In addition, BMI outliers <17.5 kg/m2 or >40 kg/m2 (GNHIES98: n = 193, 2.7 %; DEGS1: n = 215, 3 %) were excluded since the aim of the study was to describe the BP-BMI association over a BMI range that is common in the general population, while at the lowest and highest extremes of BMI a ceiling effect on BP is likely. Analyses were first run for all participants (n = 6,931 for GNHIES98 and n = 6,861 for DEGS1), then for participants without antihypertensive medication as defined above (n = 5,663 for GNHIES98 and 4,755 for DEGS1).
The average of the second and third blood pressure measurements were used for analysis. The GNHIES98 BP data were calibrated for comparison with DEGS1 data based on a formula from a methodological study described previously . In brief, the GNHIES98 and DEGS1 BP protocols were compared according to the principles of the International Protocol revision 2010 for the validation of blood pressure measuring devices in adults of the European Society of Hypertension in a measurement sequence with 105 participants yielding 315 measurement pairs. SBP and DBP values were higher with the mercury sphyghmomanometer in the GNHIES98 protocol as compared to the Datascope measurements with the DEGS1 protocol. Measurement differences increased with BP, pulse pressure, the difference in the ratio of cuff width to arm circumference, age and sex (higher mean difference in men compared to women).
The DEGS1 data were weighted to the population in Germany as of 31 Dec 2010 with respect to age, sex, region and nationality as well as type of municipality and education. The weighting factor for DEGS1 considered the re-participation probability of the former GNHIES98 participants based on a logistic regression model . The GNHIES98 data were also weighted to the population structure as of 31 Dec 2010, but to the 1998 educational distribution because of the secular changes in education levels. All analyses were weighted, so that the differences observed are controlled for age and sex.
Mean SBP, DBP and the prevalence of hypertensive BP (≥140/90 mmHg) and hypertension (BP ≥140/90 mmHg or taking ATC-coded antihypertensive medication in case of known hypertension) were assessed in both sexes and in subgroups of age, BMI, antihypertensive medical treatment, alcohol consumption, physical activity, smoking status and socioeconomic status (SES) for both surveys. Differences between the surveys were tested with chi-squared tests for categorical variables and t-tests for continuous variables. Tests were considered significant if p ≤ 0.05. To visualize the association of BMI with SBP, scatterplots of BMI and SBP were produced by combining the following two plots in one graph: 1.) BMI was divided into 5 % BMI percentile ranges (<P5, P5-P10 etc. up to P95-P100) and the mean BMI and mean SBP were plotted for these percentile ranges; 2.) curves of predicted SBP values from unadjusted linear regression models of SBP on BMI, including BMI squared (BMI2). Generalized linear regression analyses were conducted separately for GNHIES98 and DEGS1 and were stratified by sex (since a BMI x sex interaction was significant, p = 0.000) and treatment (all and untreated participants). BMI2 was included to allow for a non-linear association of BMI and SBP. The significance of the BMI-SBP-association was tested in a combined test for BMI and BMI2.
To investigate whether the BMI-SBP-association changed between 1998 and 2008–11, both surveys were combined and a BMI × survey and BMI2 × survey interaction were included in the model. The significance of the interaction was tested in a combined test for BMI × survey and BMI2 × survey. Moreover, all analyses were adjusted for selected covariates including age, antihypertensive medication, alcohol consumption, physical activity, smoking status and SES. The analyses were computed with the complex samples option in SPSS 20.0, using the LMATRIX option for the combined tests.