Objective: Recently we reported a complete relapse in the blood pressure (BP) of obese subjects despite a maintained 16% weight loss over 8 years. This relapse is now analyzed as a function of several variables. Pulse pressure (PP) is an independent risk factor of cardiovascular mortality. We now examine the development of PP in the obese and whether it can be modified by weight-reducing gastric surgery.
Research Methods and Procedures: A total of 1157 patients treated with gastric surgery and 1031 obese controls (body mass index of 41.0 ± 4.6 kg/m2 [mean ± SD], age 48 ± 6 years) were followed for 5.5 ± 2.1 (range 3 to 10) years. To separate the effect of weight change from effect of time on BP, the patients were divided in cohorts based on follow-up time.
Results: Gastric surgery resulted in a maximum weight loss after 1 year that was followed by a moderate relapse. After 5.5 years, weight loss in the intervention group was 18 ± 11% of initial body weight. Very little weight change was seen in controls. Systolic BP decreased in the intervention group during the first 6 months but had relapsed to control values at last examination. The adjusted change in PP was +4.7 mm Hg in obese controls but +2.9 mm Hg in the intervention group (p < 0.001). Final BP values were more closely related to follow-up time and ongoing weight increase than to initial body weight or initial weight loss.
Discussion: Effects of time (aging) and weight change per year on BP can be separated. An early increase in PP could be observed in the obese. This increase could be modified by weight-reducing gastric surgery.
Obesity is associated with elevated cardiovascular risk factors and increased morbidity and mortality (1) (2). Although it is generally accepted that risk factors are reduced by weight reduction (3), several observational epidemiological studies have indicated that weight reduction is also associated with increased mortality among individuals who were overweight at baseline (4).
Several explanations for the opposite effect of weight reduction on risk factors and cardiovascular mortality have been suggested. The most common interpretation has been that observational studies cannot separate intentional weight loss from unintentional weight reduction due to subclinical disease. Neither has it been possible to separate loss of fat mass from loss of lean body mass. Another cause might be that mortality data are usually long-term observations, whereas the effects on risk factors by weight reduction are almost exclusively short-term observations (3). Thus, it is not known whether initial positive effects on risk factors are persistent over time in obese individuals with maintained weight losses.
One indication in the latter direction was obtained in an orlistat trial (5) in which all risk factors responded positively and almost maximally to an initial weight reduction during single blind run-in conditions. In the placebo group, which achieved a weight reduction of 6% at the end of the first year and 4% at the end of the second year, two risk factors had relapsed to baseline (pre-run-in) values already during the first year. In contrast, all risk factors were positively influenced over 2 years in the orlistat group, which achieved 10% weight reduction at the end of the first year and 8% at the end of the second year. Thus, the degree of weight loss determines the duration of the risk factor reduction.
More compelling evidence in this direction was recently obtained from the Swedish Obese Subjects (SOS) study, which consisted of one group treated with gastric surgery and one group treated conventionally in primary healthcare settings (6). The surgically treated group maintained a 16% weight loss over 8 years. This had a dramatic effect on the incidence of diabetes. However, there was no reduction in the incidence of hypertension as compared with the conventionally treated group who did not have any longterm weight reduction (6). Thus, different risk factors can behave differently in response to maintained long-term weight reduction.
In our previous paper (6), we could not separate the effects of time (aging) from the effect of weight change per unit of time on the relapsing blood pressure (BP). One aim of this paper is to separate time (aging) from weight and weight changes as explanations of the long-term relapse from initially reduced BP values.
Another finding of our previous SOS paper (6) was that diastolic BP (DBP) was continuously falling in the conventionally treated control group but increasing in the surgically treated patients after the initial fall. Systolic BP (SBP) was increasing in both groups. In that paper (6), the difference in pulse pressure (PP) change between the two groups did not reach full significance. PP changes of surgically treated patients and controls will be revisited in this paper, with a larger number of patients included. This issue is of great interest because it has been demonstrated that a high PP is associated with an increased risk of coronary heart disease (7). In men, a high PP has also been associated with increased cardiovascular risks (8) (9) as well as overall mortality (10). In randomly selected individuals, SBP increases over the whole life span, whereas DBP reaches a maximum at ∼60 years of age (11) (12). After this age, PP increases rapidly. Given these conditions, other aims of this report are to examine whether the increase in PP can be detected earlier in life among obese individuals and whether this increase can be counteracted by weightreducing gastric surgery.
Research Methods and Procedures
The SOS study is an ongoing project consisting of a registry and an intervention study. The registry study is a health examination of obese individuals undertaken by 480 primary healthcare centers in Sweden. From the registry, eligible patients are recruited into the intervention study. This study consists of two groups: one surgically treated group and one matched, conventionally treated control group. The matching procedures are computerized, and no manual influence from the investigators is possible. The matching program considers 18 different variables: sex (absolute match), age, weight, height, waist circumference, hip circumference, SBP, s-cholesterol, s-triglycerides, smoking, diabetes, pre/postmenopausal state among women, four psychosocial variables known to be associated with mortality (current health, availability of social interaction, availability of attachment, and stressful life events), and two personality traits related to treatment preferences (psychastenia and monotony avoidance). The inclusion day of the surgically treated subject and his/her matched control is the operation day of the former. Inclusion criteria for the intervention study are as follows: age, 37 to 60 years; body mass index (BMI) of ≥38 kg/m2 for women and ≥34 kg/m2 for men. Severe illness, abuse of alcohol or drugs, and previous bariatric surgery were reasons for exclusion, whereas diabetes, hypertension, and previously experienced (not in the last 6 months) myocardial infarction were not. All details of the study design have been published previously (13).
The operations were performed at 25 surgical departments located throughout Sweden. Three types of gastric surgery were performed: vertical banded gastroplasty (VBG), gastric banding (GB), and gastric bypass (GBP) (14). It was neither feasible nor scientifically desirable to introduce a standardized treatment for the controls at the 480 participating primary healthcare centers. Instead, SOS controls received the customary obesity treatment of the site to which they belonged. No anti-obesity drugs were registered in Sweden during the study period. Given the poor long-term results after traditional obesity management (15), poor weight-loss effects were anticipated after nonsurgical treatment. Thus, these patients were expected to constitute a control group, which on average would not experience intentional weight loss.
Principal of Current Study
In the previous paper (6) mentioned in the introduction, it was not possible to separately analyze the effects of aging and weight change per time unit on the relapsing BP because the two former processes were parallel phenomena. That paper (6) used a formal intention-to-treat (ITT) analysis. This report should be considered a post hoc analysis to separate the effect of aging from the effect of weight change per unit of time as illustrated in Figure 1.
Based on follow-up, surgically treated patients and controls were divided into five different groups with 3, 4, 6, 8, or 10 years of follow-up. Those patients followed for a larger number of years were not included in the cohorts that were followed for a smaller number of years. The years of follow-up were based on backward calculations from June 12, 1999. For surgically treated patients and controls, the following independent variables were analyzed in relation to final BP: inclusion weight, weight change (usually weight loss) during the first year (Period I; Figure 1), weight change per year between the end of the first year and the second to last observation (Period II), weight change per year between the second to last observation and the last observation (Period III), and time between the start of intervention and the last observation. By using the five independent subgroups with different follow-up times, it was possible to separate the effect of weight change per year from the effect of time (aging) on final BP within the two treatment groups (surgery or conventional treatment). The second to last observation could occur 2, 3, 4, 6, or 8 years after the start of intervention, whereas the last observation could occur 3, 4, 6, 8, or 10 years after start of intervention (Figure 1).
To study the relationships between final BP and the independent variables in controls and surgically treated patients, only patients with a follow-up between 3 and 10 years who had not changed treatment assignment were used. This design was chosen because a formal ITT analysis might have distorted the true relationships between the independent variables and BP. For simplicity, results from the same set of patients have been used when comparing the effects of different treatments on final BP levels. ITT calculations (data not shown) did not change any of the conclusions.
Table 1 gives the characteristics of completers in the 10 subgroups at the start of intervention. Dropouts are defined as those patients who never completed the 3-year examination, those who shifted treatment from conventional to surgical or from surgical to conventional, and those who were converted from one surgical technique to another. This dropout frequency was 20% among surgically treated individuals and 21% among controls.
Table 1. Characteristics of controls and surgically treated at start of intervention
Values are means ± SD. Patients were divided into cohortsby follow-up time.
49 ± 6
49 ± 6
49 ± 6
49 ± 6
48 ± 6
49 ± 6
38.8 ± 4.1
39.6 ± 4.2
39.8 ± 4.5
39.3 ± 4.8
40.7 ± 4.8
39.6 ± 4.4
SBP (mm Hg)
139 ± 18
137 ± 19
138 ± 17
139 ± 18
141 ± 17
138 ± 18
DBP (mm Hg)
85 ± 10
84 ± 11
86 ± 10
87 ± 11
87 ± 11
85 ± 11
48 ± 6
47 ± 6
47 ± 6
47 ± 6
47 ± 6
47 ± 6
42.5 ± 4.6
42.8 ± 4.5
41.8 ± 3.9
41.4 ± 3.6
43.0 ± 4.6
42.3 ± 4.3
SBP (mm Hg)
144 ± 18
144 ± 17
142 ± 18
145 ± 19
147 ± 21
144 ± 18
DBP (mm Hg)
89 ± 11
90 ± 11
89 ± 11
90 ± 11
91 ± 11
90 ± 11
Matching and Baseline Characteristics
Figure 1 illustrates that the three groups to be surgically treated were initially not significantly different in body weight; however, taken together they weighed slightly more than the matched controls. The matching occurred 1.1 ± 2 years before inclusion in the intervention. While waiting for inclusion, the patients to be treated surgically increased in weight, whereas the controls decreased (Figure 1). At inclusion (i.e., start of intervention), the BMI of the surgical group was 42.3 ± 4 kg/m2 and the BMI of the control group was 39.6 ± 4 kg/m2 (p < 0.001; Table 1). At matching, neither SBP nor DBP was different between controls and patients to be surgically treated (Figure 2). Between matching and inclusion, BP changed in parallel with body weight in both groups (Figure 2); at inclusion, SBP and DBP were higher in the group to be surgically treated (Table 1). Age was 1.5 (95% confidence interval [CI]: 1.0, 2.0) years lower in the group to be surgically treated.
From a biological point of view, the division by follow-up years resulted in similar subgroups, although inclusion BMI was statistically different between some subgroups within both the surgically treated patients and the controls. In addition, sex distribution differed between some subgroups within the control group (Table 1).
The following anthropometric measurements have been used in this report. Body weight was measured to the nearest 0.1 kg (wearing indoor clothing and no shoes), and body height was measured to the nearest 0.01 m. SBP and Korotkoff phase 5 DBP were measured once after 15 minutes in a supine position. The patient spent the last 5 minutes of these 15 minutes at complete rest. Cuff width and upper arm circumference were recorded in each individual case. The BP measurements were adjusted for any incongruities between cuff size and upper arm circumference before analysis (16).
Statistical Methods and Adjustments
The statistical calculations were performed with Stata 6.0 software (StataCorp, 1999, College Station, TX). Linear regression and ANOVA according to the general linear model were used. Adjustments were made in all calculations for any differences between the groups at inclusion.
In Figure 2, the adjustments for inclusion BP levels were made from the 6-month examination and onward. Unadjusted matching values for the BP measurements were added to visualize conditions at matching. Unadjusted values are used to make it possible for the reader to estimate the effects of adjustments (see Results and Figure 2). When adjusted, PP is always adjusted for both inclusion SBP and DBP.
The study was performed according to the declaration of Helsinki and was approved by the Ethics Committees of all universities in Sweden. Patients gave their informed consent.
Changes in Body Weight and BP
The conventionally treated control group continued its small weight loss, initiated already before the start of the intervention (Figure 1), and reached, on average, a nadir at 6 months. At this time, the weight loss was −1.6 ± 6.6 kg (range, −37 to 18 kg; −1.3 ± 5.7%; range, −30% to 20%) as compared with the start of intervention. Between 6 and 12 months, no significant weight change occurred. After 1 year, a slow relapse started. At the last examination, body weight was 1.5 ± 10.2 kg (range, −30 to 47 kg; 1.6 ± 9.3%; range, −24% to 48%) above the weight at the start of the intervention.
Surgery resulted in rapid weight losses over the first 6 months and, on average, nadir weights at 1 year. At that time, the weight reductions were −25.8 ± 12.9 kg, −30.7 ± 11.8 kg, and −44.0 ± 15.0 kg (−21%, −25%, and −35%, respectively) in patients treated with GB, VBG, and GBP, respectively. From 1 year and onward, the three groups relapsed slowly. Compared with the start of intervention, the weight losses were 20.7 ± 16.6 kg, 20.8 ± 13.1 kg, and 33.8 ± 18.1 kg (17%, 17%, and 27%, respectively), respectively, at the last examination. All changes of the surgically treated groups at 1 year and at the last examination were markedly different from the changes in the conventionally treated control group (p < 0.0001).
In the control group, the unadjusted SBP dropped from 140 ± 19 mm Hg to 138 ± 18 mm Hg during the small weight loss between matching and start of intervention. Although SBP was virtually unchanged during the first year of intervention, it increased later and had reached 142 ± 18 mm Hg at the last examination. The unadjusted DBP, in contrast, decreased continuously over the total observation period from 87 ± 11 mm Hg at matching, through 85 ± 11 mm Hg at the start of intervention, to 84 ± 10 mm Hg at the last observation. Adjustments did not change these patterns of SBP and DBP (Figure 2).
In the patients to be surgically treated, unadjusted SBP increased from 140 ± 18 mm Hg to 144 ± 18 mm Hg during the increase in body weight between matching and start of intervention. During the fast weight reduction between the start of intervention and 6 months later, SBP dropped 9 ± 1 mm Hg. After 6 months, SBP started to relapse despite continuing weight loss up to 1 year (Figure 1). Unadjusted SBP was 144 ± 20 mm Hg at last examination. Unadjusted DBP increased from 88 ± 11 mm Hg to 90 ± 11 mm Hg between matching and inclusion. After an initial reduction of 6 ± 3 mm Hg, a relapse also occurred in the DBP. At the last observation, DBP had reached 87 ± 11 mm Hg. Adjustments did not change these patterns (Figure 2).
Determinants of BP Level at Last Examination
In Table 2, the final SBP and DBP of surgically treated subjects and controls had been regressed by years of follow-up, inclusion weight, weight change from start of intervention until 1-year examination (Period I), weight change per year from 1-year examination until the second to last examination (Period II), and weight change per year from the second to last examination to the last examination (Period III) (Figure 1).
Table 2. Adjusted multiple regressions of final SBP and DBP by follow-up time, inclusion weight, weight change during the first year (Period I), weight change per year from year 1 to the second to last observation (Period II), and weight change per year from the second to last to the last observation (Period III)
Mean ± SD
Final SBP mm Hg (95% CI)
Final DBP mm Hg (95% CI)
Mean ± SD
Final SBP mm Hg (95% CI)
Final DBP mm Hg (95% CI)
Values are β coefficients (95% CI). p < 0.0001 for all equations. Adjustments are made for sex, age, inclusion height, and inclusion BP, and for the following variables at the last observation: smoking (yes/no), alcohol intake (grams per day), and BP medication (yes/no).
5.4 ± 2
1.24 (0.73, 1.77)
0.54 (0.24, 0.84)
5.6 ± 2
1.25 (0.80, 1.70)
0.09 (−0.18, 0.36)
Inclusion weight (kg)
121 ± 16
0.16 (0.06, 0.27)
0.11 (0.05, 0.17)
114 ± 16
−0.02 (−0.10, 0.06)
0.00 (−0.04, 0.05)
Period I (kg/year)
−30.3 ± 13
0.14 (0.04, 0.25)
0.14 (0.08, 0.20)
−1.3 ± 8
0.23 (0.09, 0.37)
0.14 (0.07, 0.22)
Period II (kg/year)
2.4 ± 5
0.34 (0.11, 0.57)
0.27 (0.14, 0.41)
0.9 ± 4
0.29 (0.00, 0.56)
0.14 (−0.03, 0.30)
Period III (kg/year)
1.9 ± 5
0.30 (0.10, 0.51)
0.20 (0.08, 0.32)
0.6 ± 5
0.23 (0.03, 0.42)
0.17 (0.06, 0.29)
In the surgical group, an increase of 1 kg/y during Period II or Period III had approximately twice as large an impact on final SBP or final DBP as 1 kg of weight loss during the first year (Period I) or a 1 kg difference in body weight at inclusion (Table 2). Compared with 1 kg of weight change per year during Period III, 1 year of follow-up had a four times larger impact on the final SBP and a two to three times larger impact on the final DBP.
Among controls, inclusion weight was not related to final SBP or final DBP. The remaining four predictors were related to the final SBP of controls in a way similar to the situation among surgically treated subjects (Table 2). Final DBP was significantly related only to body weight changes during Periods I and III.
No significant interaction was found between the effects of age at the start of intervention and follow-up time on final BP levels, when examined by adding an interaction term to the multivariate regressions.
Treatment-Specific Changes of PP and BP
PP was increasing faster in controls than in the surgically treated group. This trend was examined by calculating changes from the start of intervention until the final observation. Adjustments were made for any incongruities between the groups (Table 3).
Table 3. Adjusted ΔPP, ΔSBP, and ΔDBP (95% CI) by treatment assignment
Values are mean changes from inclusion to last observation (95% CI). Adjustments are made for sex, age, follow-up time, inclusion weight, inclusion height, and inclusion BP, and for the following variables at the last observation: smoking (yes/no), alcohol intake (grams per day), and BP medication (yes/no).
Table 3 shows that there was no difference in ΔSBP between the control group and the surgically treated group. However, in GBP patients, SBP was reduced by 8.3 mm Hg, which was significantly different from the change in controls (+1.6 mm Hg). A decrease in DBP of 3.2 mm Hg was observed in controls. The decrease in DBP was significantly less in VBG patients (1.0 mm Hg) and GB patients (1.7 mm Hg) and significantly larger in GBP patients (6.7 mm Hg).
Because of changes in SBP and DBP, the change in PP was +4.7 mm Hg in controls. The PP increase was not different in the VBG group (+3.7 mm Hg) but was smaller among patients who had undergone GB (+1.4 mm Hg) and GBP (−1.5 mm Hg). Taken together, the increase in PP was smaller in surgically treated individuals (+2.9 mm Hg) than in controls (+ 4.7 mm Hg; p = 0.002).
Similar values with smaller increases of PP in patients treated with GB and GBP were observed if the calculations were based on matching values rather than on start of intervention values. This was true both in completer and ITT populations and independent of whether or not data were adjusted (data not shown).
This report has analyzed the relapse in BP seen after a large maintained weight loss induced by gastric surgery in severely obese subjects. The BP at the last examination was more closely related to time (aging) and ongoing weight change than to initial body weight and initial weight loss.
The beneficial effect of surgically induced weight loss on BP is well-documented by 2-year (17) and 4-year (18) (19) trials. However, in an 8-year perspective, BP levels and incidence of hypertension were not influenced by a maintained 16% weight reduction induced by gastric surgery (6).
The final BP in the operated population was not equally related to an earlier lost and a later regained amount of weight. In fact, a given weight re-gain in the later parts of the study had almost twice as large an absolute impact on final BP as a corresponding difference in inclusion weight or the same degree of weight loss in the beginning of the study. Our results are in agreement with an observational study showing that ongoing weight increase is of greater importance for current BP than is baseline body weight (20). However, neither this (20) nor other short-term (17) (21) or long-term (6) (18) (22) studies have attempted or separated the effects of time (aging) from the effect of the simultaneous weight change per unit of time on the BP level at the last observation.
In the current study, we found that in operated patients, the effect on BP of 1 elapsed year was 2.5 to 4 times larger than the effect of 1 regained kg in the later phases of the study.
No interaction effect was found between the effects of age at start of intervention and elapsed follow-up time on final SBP or DBP. However, it must be noted that in this study, the age range at the start of the intervention was between 37 and 60 years. In an older population with further developed atherosclerosis, an interaction might well have been found.
Although the effect of time on SBP was the same among controls as in the surgically treated group, no significant association between follow-up time and DBP was found in the control group. This discrepancy might be related to a slope closer to zero for DBP versus time in the obese controls than for other pressure changes observed in this study. Other deviating findings in controls, as compared with surgically treated patients, were that the inclusion weight did not predict final BP and that the small initial weight decrease had the same impact on final BP as the late weight increase. We cannot provide explanations for these differences between surgically treated individuals and controls.
Although a decrease in DBP is known to occur in normal weight populations from an age of ∼60 years (11) (12), this phenomena was observed ∼10 years earlier in weight-stable severely obese men and women (6). DBP decreases at a rate of ∼1 to 2 mm Hg per decade after the age of 60 in non-obese subjects (11). In our obese population, this rate was substantially higher: DBP decreased by 3.2 mm Hg after a mean follow-up of 5.5 years in weight-stable obese controls aged 49 years at inclusion.
The combination of a continuously increasing SBP and a decreasing DBP increases the PP. The increase in PP is seen as a consequence of an increasing arterial stiffness (23) and has been associated with an increased intima-media thickness (24). In an earlier SOS report, it was shown that the progression rate of the intima-media thickness in the carotid bulb of obese weight-stable subjects is three times faster than in age-matched lean controls (25). It was also shown that the surgically induced weight reduction normalized this progression rate (25). These observations agree with our finding that a maintained large weight reduction slows the rapid increase in PP seen in weight-stable severely obese subjects. The slower increase in PP may prove to be a beneficial effect of surgically induced weight reduction because PP has appeared as an independent predictor of coronary heart disease (7) and cardiovascular mortality (8) (9).
VBG had a less pronounced effect on PP as compared with GB and GBP. The reasons for this have not been clarified because GB and VBG resulted in similar weight losses. Whether this is a chance finding or is instead an observation related to different effects of GB and VBG on gastrointestinal signaling is an open question. The most favorable response in BP and PP was seen in GBP patients. GBP is associated with increased releases of glucagon-like peptide 1 (GLP-1) after meals, and GLP-1 has weight-reducing and anti-diabetic properties (26). Unfortunately, GLP-1 was not measured in this study.
A weakness of the SOS study is the nonrandomized design, which might introduce various biases. The chosen design was a necessity because the Ethics Committees in Sweden did not approve a randomized design of SOS. Although the unadjusted and adjusted evaluations resulted in similar conclusions, we cannot exclude the possibility that the matching procedure resulted in undetected differences between the groups.
Another possible bias is the BP measurement. Cuff sizes not adapted to an obese population may render unreliable readings. The study sites were instructed to use adequate cuff sizes, but in some instances they did not. To control for this as much as possible, arm circumferences and cuff sizes were recorded at all examinations. This made it possible to mathematically adjust for any suboptimal combinations (16).
This report suggests that the relapse in BP seen after surgically induced large weight losses in obese subjects is more related to aging and recent small weight increases than to inclusion weight or initial weight losses. It has not previously been possible to separate these effects from each other. Weight-reducing gastric surgery slows the rise in PP seen in weight-stable obese controls. This may indicate that weight loss could reduce the elevated progression rate of the atherosclerotic process observed in obese subjects.
This study was supported by the Swedish Medical Research Council (Grant 05239), Stockholm, Sweden and by Hoffman-La Roche, Basel, Switzerland. Support was also obtained from the Volvo Research Foundation, Göteborg, Sweden; Centrum för Forskning om Offentlig Sektor, Göteborg, Sweden; The Swedish Social Welfare Board, Stockholm, Sweden; Ministry of Education, Stockholm, Sweden; and Skandia Insurance, Stockholm, Sweden. We thank the members of the steering, laboratory, and safety-monitoring committees of SOS. Furthermore, we thank the staff at 25 surgical departments and 480 primary health care centers in Sweden for excellent help and cooperation.