Available online at

www.sciencedirect.com

Diabetes & Metabolism 39 (2013) 306–313

Original article

Changes in body mass index following newly diagnosed type 2 diabetes and
risk of cardiovascular mortality: A cohort study of 8486 primary-care
patients
J. Bodegard a,∗ , J. Sundström b , B. Svennblad c , C.J. Östgren d , P.M. Nilsson e , G. Johansson b
a
Department of Medicine, AstraZeneca AB, Karlebyhus, 151 85 Södertälje, Sweden
b Department of Public health and Caring Science, Uppsala University, Uppsala, Sweden
c Uppsala Clinical Research Center, Uppsala University, Uppsala, Sweden
d Department of Medical and Health Sciences, Linköping University, Linköping, Sweden
e Department of Clinical Sciences, Lund University, University Hospital, Malmö, Sweden

Received 6 February 2013; received in revised form 25 April 2013; accepted 9 May 2013

Abstract
Aims. – Elevated body mass index (BMI) is associated with an increased risk of type 2 diabetes and cardiovascular disease (CVD). This study
explored the association between BMI changes in the first 18 months of newly diagnosed type 2 diabetes and the risk of long-term CVD mortality.
Methods. – A total of 8486 patients with newly diagnosed type 2 diabetes and no previous history of CVD or cancer were identified from 84
primary-care centres in Sweden. During the first year after diagnosis, patients were grouped according to BMI change: ‘Increase’, or ≥ +1 BMI unit;
‘unchanged’, or between +1 and–1 BMI unit; and ‘decrease’, or ≤ –1 BMI unit. Associations between BMI change and CVD mortality, defined as
death from stroke, myocardial infarction or sudden death, were estimated using adjusted Cox proportional hazards models (NCT 01121315).
Results. – Baseline mean age was 60.0 years and mean BMI was 30.2 kg/m2 . Patients were followed for up to 9 years (median: 4.6 years). During
the first 18 months, 53.4% had no change in their BMI, while 32.2% decreased and 14.4% increased. Compared with patients with unchanged BMI,
those with an increased BMI had higher risks of CVD mortality (hazard ratio: 1.63, 95% CI: 1.11–2.39) and all-cause mortality (1.33, 1.01–1.76).
BMI decreases had no association with these risks compared with unchanged BMI: 1.06 (0.76–1.48) and 1.06 (0.85–1.33), respectively.
Conclusion. – Increased BMI within the first 18 months of type 2 diabetes diagnosis was associated with an increased long-term risk of CVD
mortality. However, BMI decrease did not lower the long-term risk of mortality.
© 2013 Elsevier Masson SAS. All rights reserved.

Keywords: Epidemiology; Type 2 diabetes; Weight control; Cardiovascular disease mortality

Résumé
Impact des modifications de l’IMC après le diagnostic de diabète de type 2 sur le risque à long terme de mortalité cardiovasculaire chez
8486 patients en soins primaires.
Objectif. – Un indice élevé de masse corporelle (IMC) est associé à un risque accru de diabète de type 2 et de maladies cardiovasculaires (CV).
Nous avons étudié l’association entre l’évolution de l’IMC au cours des 18 mois après le diagnostic du diabète de type 2 et le risque de mortalité
CV à long terme.
Méthodes. – Un total de 8486 patients diabétiques de type 2 nouvellement diagnostiqués et sans antécédent de cancer ou de maladies CV issus de
84 centres de soins primaires en Suède ont été étudiés. Au cours de la première année après le diagnostic, les patients ont été regroupés en fonction
de l’évolution de l’IMC (augmentation ≥ 1 unité d’IMC; « inchangé » = entre +1 et –1 unité d’IMC ; « diminution » ≤ –1 diminution unité d’IMC).
Les associations entre l’IMC et la mortalité CV, définie comme le décès par accident vasculaire cérébral, infarctus du myocarde ou mort subite,
ajustées ont été estimées par des modèles de Cox à risques proportionnels.
Résultats. – L’âge moyen à l’inclusion était de 60,0 ans et l’IMC moyen de 30,2 kg/m2 . Les patients ont été suivis pendant neuf ans (médiane
de 4,6 ans). Pendant les 18 premiers mois, 53,4 % n’ont pas changé leur IMC, 32,2 % ont eu une diminution, et 14,4 % une augmentation. Par
rapport aux patients avec un IMC inchangé, le groupe présentant une augmentation de l’IMC avait un risque plus élevé de mortalité CV (HR: 1,63

∗ Corresponding author. Tel.: +47 930 45 811.

E-mail address: [email protected] (J. Bodegard).

1262-3636/$ – see front matter © 2013 Elsevier Masson SAS. All rights reserved.
http://dx.doi.org/10.1016/j.diabet.2013.05.004

Downloaded for Anonymous User (n/a) at CK Mobile App Test from ClinicalKey.com by Elsevier on March 16, 2024.
For personal use only. No other uses without permission. Copyright ©2024. Elsevier Inc. All rights reserved.
 J. Bodegard et al. / Diabetes & Metabolism 39 (2013) 306–313 307

[IC 95 %: 1,11–2,39]) et de mortalité toutes causes (HR: 1,33 [1,01–1,76]). La diminution de l’IMC ne modifiait pas le risque de mortalité (HR:
1,06 [de 0,76– 1,48] et 1,06 [0,85–1,33]) respectivement.
Conclusion. – L’augmentation de l’IMC au cours des 18 premiers mois après le diagnostic de diabète de type 2 est associée à une augmentation
à long terme du risque de mortalité cardiovasculaire. La diminution de l’IMC ne modifie pas le risque de mortalité.
© 2013 Elsevier Masson SAS. Tous droits réservés.

Mots clés : Épidémiologie ; Diabète de type 2 ; Le contrôle du poids ; La mortalité cardiovasculaire

1. Introduction primary care in Sweden, and codes 250 (ICD 9) and E10–E14
(ICD 10) were used to identify type 2 diabetes. The Anatomi-
Weight control or the attainment of optimal body weight is a cal Therapeutic Chemical (ATC) classification system code A10
recommended treatment goal in type 2 diabetes patients based was used to identify the prescription of glucose-lowering drugs.
on the subsequent beneficial effects on cardiovascular disease Patients were excluded from the study cohort if they had a
(CVD) risk factors [1–3]. The suggested CVD risk reduction previous history of prevalent diabetes from 1987 up to the data-
with weight loss has been supported by one small observational extraction date, or were aged either less than 35 years or more
study of diabetes patients and by extrapolated data from non- than 79 years, or newly diagnosed with diabetes after 2008.
diabetic populations [4]. Bariatric surgery for severe obesity with Patients aged less than 35 were excluded to lower the risk of
sustained and substantial weight losses of 14–25% over several type 1 diabetes patients in the main data extraction from the
years has also shown significantly reduced risk of mortality com- ROSE study. The exclusion of patients aged more than 79 years
pared with untreated patients [5]. However, the Look AHEAD was decided upon because they were considered to have lit-
(Action For Health in Diabetes) trial was prematurely termi- tle benefit from weight loss and low risk of weight gain. In
nated in 2012 because it failed to demonstrate any associations addition, subjects were excluded if they had a history of CVD,
between CVD risk and sustained moderate weight loss in dia- active cancer or missing data on BMI at baseline and within 18
betic patients using prospective lifestyle interventions [6]. The months of the diabetes diagnosis (Supplementary patient flow
clinical effect of weight loss on CVD in the context of routine chart, Fig. S1). Patients were also excluded from analyses if they
lifestyle changes in general diabetes care has previously been developed CVD or cancer during the time between the two BMI
debated and remains unsettled in the light of recent results [6,7]. measurements.
Weight gain has been reported to be associated with increased BMI was mainly chosen because it reflects body composi-
CVD risk in diabetic patients; however, clinical interpretation tion (underweight, normal weight, overweight or obese), which
is difficult due to the secondary nature of the results and the determines the risk of different weight changes [14]. Changes
methods used for weight-change calculations [8,9]. Indeed, the in weight were assessed by calculating BMI changes within the
importance of weight changes in type 2 diabetes patients on fixed first 18 months of being newly diagnosed with diabetes. Base-
outcomes is currently unclear, thereby supporting the need for line BMI was defined as the closest registration found in the
new studies addressing this issue. interval from 15 months before to 45 days after the time of
The objective of the present study was to investigate the newly diagnosed diabetes. The second BMI was sought in the
association between weight change and risk of CVD mortal- interval from 46 days to 18 months after newly diagnosed dia-
ity in a large primary-care-based sample of patients with newly betes, using only the last registration if several readings were
diagnosed diabetes in a real-world setting. found. BMI change was calculated by subtracting the baseline
BMI from the second BMI, after which the patients were divided
into three BMI groups: increased, or more than or equal to +1
2. Methods BMI unit; unchanged, or +1 to–1 BMI unit; and decreased, or
less than or equal to –1 BMI unit. The mean interval duration
The study was based on the Retrospective Epidemiological between the index and second BMI in the increased, unchanged
Study to Investigate Outcome and Mortality with Glucose- and decreased groups was 1.1, 1.0 and 1.0 year, respectively.
lowering Drug Treatment in Primary Care (ROSE) study [10].
In 2010, patients’ data were extracted from 84 primary-care
centres in Sweden, using the Pygargus Customized Extraction 2.1. Outcome and covariates
Program (CXP) [11], to constitute a representative sample of
both publicly and privately owned primary-care centres (61% Follow-up time started after the second BMI measurement,
and 39%, respectively) [12,13]. The 84 centres selected made and all patients were followed until death or the last date
up approximately 8% of the total number of primary-care cen- for extracting data from registers. The primary endpoint CVD
tres in Sweden. All data were extracted for the 58,326 patients mortality included all primary causes of death diagnosed with
diagnosed with type 2 diabetes between 1999 and 2009. The ICD-10 codes I00–I99. Non-fatal CVD comprised myocardial
diagnosis was identified by the registered diagnostic code and/or infarction (ICD-9 code 410; ICD-10 code I21), unstable angina
prescription of any blood glucose-lowering drug [10]. The Inter- (ICD-9 411; ICD-10 I20.0), heart failure (ICD-9 428; ICD-10
national Statistical Classification of Diseases and Related Health I11.0, I50), atrial fibrillation (ICD-9 427; ICD-10 I48), hae-
Problems (ICD) is used to classify diagnoses in both hospital and morrhagic and embolic stroke (ICD-9 430–438; ICD-10 I60,

Downloaded for Anonymous User (n/a) at CK Mobile App Test from ClinicalKey.com by Elsevier on March 16, 2024.
For personal use only. No other uses without permission. Copyright ©2024. Elsevier Inc. All rights reserved.
308 J. Bodegard et al. / Diabetes & Metabolism 39 (2013) 306–313

I61, I63.0–I63.5, I63.8–I63.9, I64), transient ischaemic attack with a smoking prevalence of 15% in the unchanged BMI group,
(ICD-10 G45) and peripheral artery disease (ICD-9 440, 441, based on the reported characteristics of Swedish type 2 diabetes
444; ICD-10 I70–I79). ICD codes were retrieved by linking the patients, which were set at 15% [9,18,19]. R statistics package
patients’ personal identification numbers to data from validated software, version 2.13.1, was used for all analyses.
national hospitalization and cause-of-death registries [15]. Data
on educational levels and marital status were extracted from 3. Results
Statistics Sweden databases.
Baseline covariates were calculated as the means of the last Baseline mean age was 60.0 years (range 35–79 years) and
three measures within 15 months before and 14 days after the mean BMI was 30.2 kg/m2 (range 16.7–58.5 kg/m2 ). Mean over-
date of diabetes diagnosis. Descriptions of temporal covariate all time between baseline and second BMI registration dates
changes were subsequently updated in the follow-up period by was 383 days (range 46–545, SD 121 days). Patients were
calculating the means of all measures within the subsequent followed for up to 9 years, with a median follow-up time of
time period from day 15 to day 365. For each 12-month period 4.6 years and 38,300 patient-years. Slightly more than half the
thereafter, the means of all measures within the time interval patients had an unchanged BMI (53.4%), and more patients had a
were calculated. decrease (32.2%) than increase (14.4%) in BMI (Table 1). Dur-
ing follow-up, the incidence rates (per 1000 patient-years) of
2.2. Statistical methods CVD mortality were 8.59 (6.35–11.38), 5.98 (4.97–7.14) and
5.74 (4.48–7.26) in the increased, unchanged and decreased
Differences in baseline data between groups (using BMI groups, respectively. Very few patients (4.1%) had less
unchanged BMI as the reference group) were tested by Student’s than 120 days between baseline and second BMI measurements.
t test and Pearson’s chi-square test according to the type of data. Patients with increased BMI were younger, with higher lev-
P values for continuous data were two-tailed, and differences els of education, lower HbA1c , higher high-density lipoprotein
between groups were considered significant if P < 0.05. CVD (HDL), lower creatinine levels, higher estimated glomerular fil-
mortality risks were estimated using Cox regression models tration rate (GFR) and less angina pectoris compared with the
selected to adjust for risk of CVD: model A = age, gender, BMI unchanged BMI group (Table 1). Patients who decreased their
at baseline and previous angina pectoris; and model B = model BMI were more often women, and had higher education levels,
A plus education, marital status and use of glucose-lowering higher baseline BMI, higher levels of triglycerides, higher blood
drugs. Directed acyclic graphs were used to minimize the risk pressure, lower creatinine levels and lower incidence of angina
of bias and identify the primary adjustment model, which led compared with the unchanged BMI group. A history of angina
to model B [16]. Formal statistical interactions between base- was less prevalent in both the increased and decreased groups
line BMI and BMI change, and between treatment group and than in the unchanged group.
BMI change, were performed using a logistic likelihood test During follow-up, both the increased and decreased BMI
comparing the original and adjusted model B. groups maintained higher BMIs compared with the unchanged
group, and the increased group had the highest readings
2.3. Sensitivity analyses (Supplementary follow-up data, Fig. S2). The overall BMI trend
was a slight decrease in BMI during follow-up in all groups. A
To study the effects of baseline BMI, patients with BMI tendency towards higher levels of HDL cholesterol and lower
≥ 30 kg/m2 were excluded. The cohort was also divided into levels of triglycerides was observed in the decreased BMI group
four groups according to baseline BMI quartiles, with survival compared with the increased group during follow-up.
analyses performed within each group. The effect of different
time intervals between baseline and second BMI was studied in 3.1. Glucose-lowering drugs
separate analyses by excluding all patients with < 120-day inter-
vals and adjusting for the time difference. The effect of time Approximately one-third of patients were taking some kind
from baseline BMI to date of diabetes diagnosis was assessed of glucose-lowering drug at the time of diabetes diagnosis
by including only patients with BMI registrations close to the (Supplementary baseline data, Table S1). During the first year,
diagnosis date (within –30 to +15 days). more patients received treatment with glucose-lowering drugs
A previous diagnosis of type 2 diabetes might have been in the increased BMI group than in the other two groups. Fewer
missed because patients had moved from one primary-care patients were using metformin in the increased group and more
centre not included for data extraction and had no hospital patients were taking it in the decreased group compared with the
registration in Sweden. To minimize the risk of previous and unchanged group. The average daily starting dose of metformin
undetected diagnoses, two separate analyses were performed by was ∼1000 mg and the overall use of metformin during follow-
excluding patients with a primary-care history < 15 months and up was similar for all three groups (Supplementary follow-up
those taking insulin treatment at baseline. data, Fig. S3).
To study the effects of an unmeasured binary confounder More patients were taking sulphonylurea and insulin in the
such as smoking on CVD mortality risk, the prevalence differ- increased BMI group and fewer in the decreased group compared
ences between the increased and unchanged BMI groups were with the unchanged group (Supplementary baseline data, Table
calculated [17]. The hazard ratio (HR) for smoking was set at 2 S1). The use of sulphonylurea and insulin was also higher during

Downloaded for Anonymous User (n/a) at CK Mobile App Test from ClinicalKey.com by Elsevier on March 16, 2024.
For personal use only. No other uses without permission. Copyright ©2024. Elsevier Inc. All rights reserved.
 J. Bodegard et al. / Diabetes & Metabolism 39 (2013) 306–313 309

Table 1
Characteristics of 8486 patients with newly diagnosed diabetes.
Increased BMI Unchanged BMI Decreased BMI
(n = 1238) (n = 4523) (n = 2725)

Age, years 57.9 (10.2)*** 59.7 (10.1) 59.0 (10.3)

Men, n (%) 720 (58.2) 2671 (59.1) 1279 (46.9)***
First BMI, kg/m2 30.1 (5.8) 30.2 (5.1) 32.7 (5.8)***
Second BMI, kg/m2 32.2 (6.0)*** 30.1 (5.1) 30.2 (5.5)
BMI change, kg/m2 2.1 (1.5)*** –0.1 (0.6) –2.5 (1.7)***
HbA1c , % 6.55 (0.80)*** 6.73 (0.81) 6.94 (0.83)
Cholesterol, mmol/L 5.64 (1.19) 5.62 (1.13) 5.68 (1.22)
LDL, mmol/L 3.45 (0.96) 3.47 (0.91) 3.50 (0.97)
HDL, mmol/L 1.47 (0.88)* 1.40 (0.78) 1.36 (0.77)
Triglycerides, mmol/L 2.26 (2.06) 2.17 (1.57) 2.33 (2.00)**
Systolic BP, mmHg 145.4 (18.6) 140.8 (17.9) 147.4 (17.9)**
Diastolic BP, mmHg 83.8 (9.4) 83.4 (9.4) 84.7 (9.3)***
Creatinine, mmol/L 78.0 (18.3)* 79.6 (17.1) 78.3 (19.0)**
Estimated GFR, mL/min 85.8 (17.5)*** 83.4 (16.8) 83.0 (17.4)
Angina pectoris, n (%) 40 (3.2)** 236 (5.2) 104 (3.8)**
COPD, n (%) 32 (2.6) 80 (1.8) 64 (2.3)
Higher education,a n (%) 769 (64.1)** 2650 (59.6) 1691 (63.1)**
Marital status
Married, n (%) 636 (51.7)*** 2637 (58.4) 1498 (55.1)*
Divorced, n (%) 232 (18.9) 765 (17.0) 496 (18.2)
Unmarried, n (%) 264 (21.5)** 748 (16.8) 504 (18.5)
Widow/widower, n (%) 97 (7.9) 363 (8.0) 222 (8.2)

BMI: body mass index; LDL: low-density lipoprotein; HDL: high-density lipoprotein; BP: blood pressure; GFR: glomerular filtration rate; COPD: chronic obstructive
pulmonary disease.
Numbers in parentheses are standard deviations (SD).
a Secondary school, college or other graduate degree.
* P < 0.05 versus unchanged BMI.
** P < 0.01 versus unchanged BMI.
*** P < 0.001 versus unchanged BMI.

follow-up in the increased BMI group and lower in the decreased use of statins, antihypertensives and low-dose ASA was similar
BMI group during the follow-up period than in the unchanged in all groups during follow-up (Supplementary follow-up data,
BMI group (Supplementary follow-up data, Fig. S3). Fig. S3), and no differences were noted in cholesterol and blood
Also, as shown in Supplementary follow-up data, Fig. S2, pressure levels across all three groups (Supplementary follow-up
from the time of being newly diagnosed with diabetes, HbA1c data, Fig. S2).
levels decreased substantially in the decreased BMI group and
remained low compared with the other two groups during follow-
up. The general trend in all three groups was, however, a slight 3.3. BMI changes associated with CVD and all-cause
increase in HbA1c levels during the follow-up period. mortality

3.2. CVD prevention drugs The risks of CVD and all-cause mortality were 83% (HR:
1.83, 95% CI: 1.29–2.58) and 48% (HR: 1.48, 95% CI:
Fewer patients in the increased BMI group were treated 1.16–1.90) higher in the patients with increased BMI compared
with acetylsalicylic acid (ASA) and antihypertensives than in with unchanged BMI, respectively (Table 2), whereas mortality
the unchanged group (Supplementary baseline data, Table S1). risk did not differ in the decreased BMI group compared with
The decreased BMI group had a slightly lower use of ASA and the unchanged group. In model B, patients with increased BMI
similar use of antihypertensives compared with the unchanged had significantly higher risks of CVD and all-cause mortality
group. However, there was greater use of thiazide diuretics in than patients with unchanged BMI at 63% (HR: 1.63, 95% CI
the decreased BMI group than in the two other groups. The 1.11–2.39) and 34% (HR: 1.34, 95% CI 1.01–1.76), respectively.

Downloaded for Anonymous User (n/a) at CK Mobile App Test from ClinicalKey.com by Elsevier on March 16, 2024.
For personal use only. No other uses without permission. Copyright ©2024. Elsevier Inc. All rights reserved.
310 J. Bodegard et al. / Diabetes & Metabolism 39 (2013) 306–313

Table 2
Risk of various endpoint events as determined by Cox model analyses.
Number of events Increased BMI Unchanged BMI Decreased BMI
(n = 1238) (n = 4523) (n = 2725)

CVD mortalitya
Age-adjusted 228 1.75 (1.24–2.48) 1.00 Reference 1.01 (0.75–1.37)
Baseline-BMI adjusted 228 1.45 (1.03–2.04) 1.00 Reference 0.98 (0.72–1.34)
Age- and baseline-BMI adjusted 228 1.77 (1.25–2.50) 1.00 Reference 0.98 (0.70–1.30)
Model Ab adjusted 228 1.83 (1.29–2.58) 1.00 Reference 1.05 (0.77–1.43)
Model Bc adjusted 197 1.63 (1.11–2.39) 1.00 Reference 1.06 (0.76–1.48)
All-cause mortality
Age-adjusted 486 1.45 (1.13–1.85) 1.00 Reference 1.04 (0.85–1.27)
Baseline-BMI adjusted 486 1.22 (0.95–1.57) 1.00 Reference 1.03 (0.84–1.27)
Age- and baseline-BMI adjusted 486 1.45 (1.13–1.86) 1.00 Reference 1.01 (0.82–1.24)
Model A adjusted 486 1.48 (1.16–1.90) 1.00 Reference 1.07 (0.87–1.32)
Model B adjusted 423 1.34 (1.01–1.76) 1.00 Reference 1.06 (0.85–1.33)

BMI: body mass index; CVD: cardiovascular disease.

a All deaths from all diagnoses involving the cardiovascular system, such as myocardial infarction, arrhythmias, stroke, heart failure, peripheral artery disease.
b Age, gender, BMI at baseline, previous angina pectoris.
c Model A + education, marital status, use of glucose-lowering drugs.

Introducing adjustments for age in the survival model changed BMI quartiles 1 to 4, CVD mortality risk was 1.73 (95% CI:
estimates of CVD and all-cause mortality the most. The adjusted 0.92–3.24), 1.20 (95% CI: 0.49–2.99), 2.12 (95% CI: 0.96–4.66)
increased cumulative risk of CVD mortality in the increased and 1.33 (95% CI: 0.52–3.39), respectively, compared with the
BMI group was evident and statistically verifiable from 3 years unchanged BMI group.
onwards (Fig. 1). For 8135 patients (95.9%), the time between baseline and sec-
ond BMI measurements was at least 120 days, and the risk of
3.4. Sensitivity analyses CVD mortality was 1.52 (95% CI: 1.02–2.27) and 1.05 (95%
CI: 0.75–1.48) in the increased and decreased BMI groups,
In the subgroup of 4458 patients (52.5%) with BMI ≥ respectively. The main results did not change with additional
30 kg/m2 , the incidence rates of CVD mortality were 7.87 adjustments for time between baseline and second BMI: 1.62
(4.83–12.17), 5.61 (4.23–7.32) and 5.84 (4.30–7.76) in the (95% CI: 1.10–2.39). When using only the BMIs registered from
increased, unchanged and decreased BMI groups, respectively. 30 days before to 14 days after the date of diabetes diagnosis,
Compared with the unchanged BMI group, the corresponding the corresponding risks of CVD mortality were 1.66 (95% CI:
risks of CVD mortality were 1.73 (95% CI: 0.98–3.07) and 1.04–2.65) and 1.15 (95% CI: 0.78–1.68) in the increased and
1.12 (95% CI: 0.73–1.72) in the increased and decreased BMI decreased BMI groups, respectively.
groups, respectively. In the subbroups according to baseline On analyzing the 7169 patients (84.5%) with a treatment his-
tory > 15 months at their respective primary-care centres, the
associations were slightly stronger than in the primary analy-
sis sample, with HRs of 1.74 (95% CI: 1.16–2.61) for CVD
mortality and 1.30 (95% CI: 0.96–1.74) for all-cause mortal-
ity in the increased BMI group versus the unchanged BMI
group. Excluding patients using insulin at baseline or within
the first year of diagnosis did not change risk estimates for CVD
and all-cause mortality in the increased BMI group versus the
unchanged group: 1.68 (95% CI: 1.12–2.52) versus 1.29 (95%
CI: 0.96–1.74), respectively.
On applying all adjusted covariates at the time of the second
BMI, the risk of CVD mortality did not change significantly
compared with the main results (1.57, 95% CI: 1.04–2.37 versus
1.63, 95% CI: 1.11–2.39). Compared with all other covariates,
BMI influenced the results the most.
On assessing the effect of smoking, it was calculated that
15% and 10% higher smoking prevalences in the increased and
unchanged BMI groups changed CVD mortality risk to 1.44
Fig. 1. Cumulative risk of cardiovascular death adjusted for age, gender, body
(95% CI: 0.98–2.11) and 1.50 (95% CI: 1.02–2.20), respectively.
mass index (BMI) at baseline, previous angina pectoris, education level, marital On introducing the product of (BMI change * baseline BMI)
status and use of glucose-lowering drugs. and (BMI change * glucose-lowering drug [sulphonylurea or

Downloaded for Anonymous User (n/a) at CK Mobile App Test from ClinicalKey.com by Elsevier on March 16, 2024.
For personal use only. No other uses without permission. Copyright ©2024. Elsevier Inc. All rights reserved.
 J. Bodegard et al. / Diabetes & Metabolism 39 (2013) 306–313 311

metformin]) in the Cox model, no interaction was detected; P of beta-cell failure than in the unchanged BMI group. The patho-
values were 0.59 and 0.22, respectively. physiology that accompanies newly diagnosed diabetes changes
clinical management, thereby introducing an unknown con-
4. Discussion founding effect. However, on excluding all non-obese patients
(BMI < 30 kg/m2 ) and those treated with insulin, thereby lower-
The present study demonstrates that slightly more than half of ing the probability of pancreatic insufficiency, the risk of CVD
patients (53.4%) maintain their BMI with no changes during the mortality changed only slightly. The associations between BMI
first 18 months after type 2 diabetes diagnosis. The least common increase and CVD mortality risk were also estimated according
observation was an increased BMI (14.4%) by more than 1 BMI to baseline BMI categories (quartiles). HRs remained numeri-
unit (∼3.6 kg), and this group had a 63% greater risk of CVD cally similar, but lost their statistical significance, and no trends
mortality compared with patients with unchanged BMI. Inter- were visible. These sensitivity analyses address some of the con-
estingly, no risk reduction was detected in the 32.2% of patients founding properties of different pathways and support the main
whose BMI decreased by more than 1 BMI unit compared with results of our study.
those whose BMI remained unchanged. These findings were As with other studies [6,32–35], our study failed to demon-
robust and remained virtually unchanged after adjusting for sev- strate any CVD risk-altering effects with weight reduction,
eral important risk factors. Furthermore, on replacing the CVD despite showing positive long-lasting effects on BMI, HbA1c ,
mortality with the all-cause mortality endpoint, the results were HDL and triglycerides (Supplementary data, Fig. S2). Recently,
still similar, albeit slightly weaker. the Look AHEAD trial was stopped by the study’s data and
Very few studies have addressed the relationship between safety monitoring board when it became evident that, although
weight change, closely associated with BMI, and incidence of the intensive lifestyle interventions to reduce weight in obese
CVD in diabetes patients. Drawing from the Swedish National type 2 diabetes patients did no harm, they also did not lower the
Diabetes Register, Eeg-Olofsson et al. [9] reported an increased risk of CVD events either [6]. However, the Look AHEAD find-
risk of non-fatal and fatal CVD with increasing weight in 4916 ings have yet to be published, making any interpretation difficult
patients with prevalent diabetes (24,144 person-years). In that at this time.
study, the assessment of weight change was a secondary objec- Other studies have reported lower risks of clinical events
tive and the method used differed substantially from ours by with weight loss. The Swedish Obese Subjects (SOS) study
calculating the BMI difference up to a CVD event or the study reported reduced mortality with substantial and sustained weight
end. On comparing patients with very high increases (median loss (14–25%) after gastric surgery [5]. Intentional weight
+3.8 BMI units) versus moderate weight increases (median +1 loss was associated with reduced CVD risk in an observa-
BMI unit), the former group was associated with a 1.54-fold tional study based on patient-reported weight changes [36].
increased risk of CVD compared with the latter. As in our In a CVD population with or without diabetes, early weight
present study, those authors observed no association between loss (over 6 weeks) of just over 2 kg lowered the risk of
weight reduction and CVD risk. However, despite considerable CVD mortality over the subsequent 5-year follow-up [37]. This
differences in methods and materials, the two studies both have suggests that, despite large differences in patients’ character-
similar findings of an increased CVD risk with increasing BMI istics and study design, the results may yet support beneficial
in patients with type 2 diabetes. effects with substantial or moderate weight loss in specified
Weight loss may be unintentional, such as that caused by subgroups of diabetes patients, such as those with established
poorly controlled diabetes or other advanced disease, or inten- CVD.
tional, such as with lifestyle modifications. Weight increase The proportion of patients with unintentional weight loss
may be due to failure of lifestyle modifications or effects of caused by poorly controlled blood glucose levels may have been
pharmacological treatment. Undesirable weight gain is related higher in the decreased BMI group (who were heavier, with
to glucose-lowering drugs that directly or indirectly increase higher HbA1c levels and less use of glucose-lowering drugs)
levels of insulin, including insulin itself [20–22] and insulin than in the unchanged BMI group. Interestingly, the slightly
secretagogues (sulphonylureas) [23–26]. In our present study, increased use of diuretics (thiazides) in the decreased BMI group
a larger proportion of patients in the increased BMI group might partially explain the unintentional weight loss while at the
were already using sulphonylurea and insulin, and more patients same time indicating low-intensity antihypertensive treatment.
added these drugs during follow-up (Supplementary follow-up Such reasons for unintentional weight loss could somewhat
data, Fig. S3), thus partially explaining the increase in BMI in compromise CVD risk estimates for the decreased BMI group.
this group (Supplementary follow-up data, Fig. S2). The pro- However, it cannot be ruled out that our findings may hold
portion of patients taking metformin was significantly lower true for BMI decreases, supported by recent results from the
in the increased BMI group and higher in the decreased BMI prospective Look AHEAD trial [6].
group both at baseline and during follow-up. These substan- The time between BMI measurements may have influenced
tial differences might also partly explain the weight changes, our results, especially considering that these were records from
supported by the notion of metformin’s weight-reducing effects real-life clinical settings. By introducing the criterion of at
[27–31]. least 120 days between baseline and second BMI measure-
Baseline characteristics of the increased BMI group (higher ments, and adjusting for time between the two measurements,
insulin use and lower HbA1c ) could indicate a greater probability the results did not change significantly and, thus, support our

Downloaded for Anonymous User (n/a) at CK Mobile App Test from ClinicalKey.com by Elsevier on March 16, 2024.
For personal use only. No other uses without permission. Copyright ©2024. Elsevier Inc. All rights reserved.
312 J. Bodegard et al. / Diabetes & Metabolism 39 (2013) 306–313

primary results. The choice of adjusting for either the index or Disclosure of interest
second BMI also had no effect on our main results, thereby
bolstering the view that BMI changes can affect future CVD J.B. holds a full-time position at AstraZeneca as an epi-
risk. demiologist. J.S., P.N., C.J.Ö, B.S. and G.J. have received
compensation for the work on this report from AstraZeneca.
Contributions: J.B. researched data, contributed to the discus-
4.1. Strengths and limitations
sion, and wrote, reviewed and edited the manuscript. B.S. had
full access to the data and performed statistical analyses. J.S.,
The strengths of the present study include the selection of
C.J.Ö, P.N., B.S. and G.J. researched data, contributed to the
patients with newly diagnosed diabetes and large numbers of
discussion, reviewed and edited the manuscript. J.B. is the guar-
patients with long-term follow-ups in real-life clinical settings
antor of this work and, as such, takes responsibility for the
through the use of nationwide healthcare registries. However,
integrity of the data and accuracy of the data analysis.
data on beta-cell function, diet, waist circumference, physical
activity and smoking were not adequately recorded in these
registries. In type 2 diabetes patients, neither smoking his- Acknowledgements
tory nor current smoking at baseline has been associated with
weight gain or weight loss, respectively [14,38]. These results Special thanks go to Professor Jan Cederholm, Department of
would argue for potentially fewer smokers in the increased Public Health and Caring Sciences/Family Medicine and Clini-
BMI group and more smokers in the decreased BMI group cal Epidemiology, Uppsala, Sweden, for advice and input to this
compared with the unchanged BMI group. There is also the work.
greater possibility of higher numbers of successful smoking Lena Ferntoft working for AstraZeneca made significant con-
cessation in the increased BMI group, given that stopping smok- tributions to this study. The authors acknowledge the database
ing is generally associated with a weight increase of 4–5 kg management by Ulf Hellström and data extraction by Pygargus
in 80% of patients over a 12-month period [39]. However, AB.
smoking interventions in a real-world setting among diabetes Funding: This study was funded by AstraZeneca.
patients have had rather disappointing results, ranging from
7% in Sweden to 2.3% in Spain [40,41]. This reasoning there- Appendix A. Supplementary data
fore suggests the possibility of a higher prevalence of smokers
in the unchanged and/or decreased BMI groups. However, the Supplementary materials (Figs. S1–S3 and Table S1) asso-
generally low prevalence of smokers and low success rates of ciated with this article can be found at http://www.scincedirect.
smoking cessation also point to less of an impact of smoking com at http://dx.doi.org/10.1016/j.diabet.2013.05.004.
behaviour on CVD risk in the increased BMI group. In addi-
tion, the estimated 10–15% difference in smoking prevalence References
applied to blunt the main results between the increased and
unchanged BMI groups seems unlikely in a country with an [1] Unick JL, Beavers D, Jakicic JM, Kitabchi AE, Knowler WC, Wadden TA,
overall 15% smoking prevalence. Moreover, the directed acyclic et al. Effectiveness of lifestyle interventions for individuals with severe
graphs used to create optimal models for minimizing bias did obesity and type 2 diabetes: results from the Look AHEAD trial. Diabetes
Care 2011;34:2152–7.
not identify smoking as a crucial confounder, which may be [2] Anderson JW, Kendall CW, Jenkins DJ. Importance of weight management
explained by the smoking-associated risk being accounted for by in type 2 diabetes: review with meta-analysis of clinical studies. J Am Coll
other covariates extracted from the otherwise detailed database Nutr 2003;22:331–9.
[16]. [3] Inzucchi SE, Bergenstal RM, Buse JB, Diamant M, Ferrannini E, Nauck M,
Furthermore, if the increased BMI group had a higher CVD et al. Management of hyperglycemia in type 2 diabetes: a patient-centered
approach: Position statement of the American Diabetes Association (ADA)
risk profile at baseline, any differences would have been appar- and the European Association for the Study of Diabetes (EASD). Diabetes
ent earlier in the follow-up. Yet, the risk difference in CVD Care 2012;35:1364–79.
mortality only became evident 3 years after the BMI changes, [4] Lean ME, Powrie JK, Anderson AS, Garthwaite PH. Obesity, weight loss
thus supporting similar risk profiles at baseline but with different and prognosis in type 2 diabetes. Diabet Med 1990;7:228–33.
rates of disease progression. [5] Sjostrom L, Narbro K, Sjostrom CD, Karason K, Larsson B, Wedel H, et al.
Effects of bariatric surgery on mortality in Swedish obese subjects. N Engl
J Med 2007;357:741–52.
5. Conclusion [6] Reiter A. Weight loss does not lower heart disease risk from type 2 diabetes.
U.S.: Department of Health and Human Services; 2012 [Available from:
http://www.nih.gov/news/health/oct2012/niddk-19.htm].
Weight gain in patients with newly diagnosed type 2 dia- [7] Nilsson PM. Is weight loss beneficial for reduction of morbidity and mortal-
betes may be more hazardous than previously recognized, and ity? What is the controversy about? Diabetes Care 2008;Suppl. 2:S278–83.
efforts should be made to prevent weight increases in diabetes [8] Doehner W, Erdmann E, Cairns R, Clark AL, Dormandy JA, Ferrannini E,
et al. Inverse relation of body weight and weight change with mortality and
patients. However, weight loss did not lower the long-term risk
morbidity in patients with type 2 diabetes and cardiovascular co-morbidity:
of either CVD or all-cause mortality. Nevertheless, the results an analysis of the PROactive study population. Int J Cardiol 2011.
of this retrospective study need to be confirmed in prospective [9] Eeg-Olofsson K, Cederholm J, Nilsson PM, Zethelius B, Nunez L, Gud-
studies before any definitive conclusions can be drawn. bjornsdottir S, et al. Risk of cardiovascular disease and mortality in

Downloaded for Anonymous User (n/a) at CK Mobile App Test from ClinicalKey.com by Elsevier on March 16, 2024.
For personal use only. No other uses without permission. Copyright ©2024. Elsevier Inc. All rights reserved.
 J. Bodegard et al. / Diabetes & Metabolism 39 (2013) 306–313 313

overweight and obese patients with type 2 diabetes: an observational study gliclazide MR and glimepiride in type 2 diabetic patients. Eur J Clin Invest
in 13,087 patients. Diabetologia 2009;52:65–73. 2004;34:535–42.
[10] Ostgren CJ, Sundstrom J, Svennblad B, Lohm L, Nilsson PM, Johansson [27] Chiasson JL, Naditch L. The synergistic effect of miglitol plus metformin
G. Associations of HbA and educational level with risk of cardiovascular combination therapy in the treatment of type 2 diabetes. Diabetes Care
events in 32 871 drug-treated patients with type 2 diabetes: a cohort study 2001;24:989–94.
in primary care. Diabet Med 2013. [28] Hanefeld M, Brunetti P, Schernthaner GH, Matthews DR, Charbonnel BH.
[11] Lidman B, Linder R. http://www.pygargus.se/3 en.html One-year glycemic control with a sulfonylurea plus pioglitazone versus a
[12] Martinell M, Stalhammar J, Hallqvist J. Automated data extraction – a fea- sulfonylurea plus metformin in patients with type 2 diabetes. Diabetes Care
sible way to construct patient registers of primary care utilization. Ups J 2004;27:141–7.
Med Sci 2012;117:52–6. [29] Pavo I, Jermendy G, Varkonyi TT, Kerenyi Z, Gyimesi A, Shoustov S, et al.
[13] Kjeldsen SE, Stalhammar J, Hasvold P, Bodegard J, Olsson U, Russell D. Effect of pioglitazone compared with metformin on glycemic control and
Effects of losartan versus candesartan in reducing cardiovascular events in indicators of insulin sensitivity in recently diagnosed patients with type 2
the primary treatment of hypertension. J Hum Hypertens 2010;24:263–73. diabetes. J Clin Endocrinol Metab 2003;88:1637–45.
[14] Fonseca V, McDuffie R, Calles J, Cohen RM, Feeney P, Feinglos M, et al. [30] Schernthaner G, Matthews DR, Charbonnel B, Hanefeld M, Brunetti P.
Determinants of weight gain in the action to control cardiovascular risk in Efficacy and safety of pioglitazone versus metformin in patients with type
diabetes trial. Diabetes Care 2013. 2 diabetes mellitus: a double-blind, randomized trial. J Clin Endocrinol
[15] Ludvigsson JF, Andersson E, Ekbom A, Feychting M, Kim JL, Reuterwall Metab 2004;89:6068–76.
C, et al. External review and validation of the Swedish national inpatient [31] Long-term safety, tolerability, and weight loss associated with metformin
register. BMC Public Health 2011;11:450. in the Diabetes Prevention Program Outcomes Study. Diabetes Care
[16] Shrier I, Platt RW. Reducing bias through directed acyclic graphs. BMC 2012;35:731–7.
Medical Res Methodol 2008;8:70. [32] Droyvold WB, Lund Nilsen TI, Lydersen S, Midthjell K, Nilsson PM,
[17] Lin DY, Psaty BM, Kronmal RA. Assessing the sensitivity of regression Nilsson JA, et al. Weight change and mortality: the Nord-Trondelag Health
results to unmeasured confounders in observational studies. Biometrics Study. J Intern Med 2005;257:338–45.
1998;54:948–63. [33] Mikkelsen KL, Heitmann BL, Keiding N, Sorensen TI. Independent effects
[18] Ekstrom N, Miftaraj M, Svensson AM, Andersson Sundell K, Cederholm of stable and changing body weight on total mortality. Epidemiology
J, Zethelius B, et al. Glucose-lowering treatment and clinical results in 163 1999;10:671–8.
121 patients with type 2 diabetes: an observational study from the Swedish [34] Nilsson PM, Nilsson JA, Hedblad B, Berglund G, Lindgarde F. The enigma
national diabetes register. Diabetes Obes Metab 2012;14:717–26. of increased non-cancer mortality after weight loss in healthy men who are
[19] Social TO. Tobacco. Brussels: Directorate General Health and Consumers; overweight or obese. J Intern Med 2002;252:70–8.
2010. [35] Sorensen TI, Rissanen A, Korkeila M, Kaprio J. Intention to lose weight,
[20] Goudswaard AN, Furlong NJ, Rutten GE, Stolk RP, Valk GD. Insulin weight changes, and 18-y mortality in overweight individuals without
monotherapy versus combinations of insulin with oral hypoglycaemic co-morbidities. PLoS Med 2005;2:e171.
agents in patients with type 2 diabetes mellitus. Cochrane Database Syst [36] Williamson DF, Thompson TJ, Thun M, Flanders D, Pamuk E, Byers T.
Rev 2004:CD003418. Intentional weight loss and mortality among overweight individuals with
[21] Jacober SJ, Scism-Bacon JL, Zagar AJ. A comparison of intensive mix- diabetes. Diabetes Care 2000;23:1499–504.
ture therapy with basal insulin therapy in insulin-naive patients with [37] Caterson ID, Finer N, Coutinho W, Van Gaal LF, Maggioni AP,
type 2 diabetes receiving oral antidiabetes agents. Diabetes Obes Metab Torp-Pedersen C, et al. Maintained intentional weight loss reduces
2006;8:448–55. cardiovascular outcomes: results from the Sibutramine Cardiovascular
[22] Balena R, Hensley IE, Miller S, Barnett AH. Combination therapy with OUTcomes (SCOUT) trial. Diabet Obes Metab 2012;14:523–30.
GLP-1 receptor agonists and basal insulin: a systematic review of the [38] van Dieren S, Czernichow S, Chalmers J, Kengne AP, de Galan BE, Poulter
literature. Diabetes Obes Metab 2012. N, et al. Weight changes and their predictors amongst 11 140 patients
[23] Luis Bautista J, Bugos C, Dirnberger G, Atherton T. Efficacy and with type 2 diabetes in the ADVANCE trial. Diabet Obes Metab 2012;14:
safety profile of glimepiride in Mexican American Patients with type 464–9.
2 diabetes mellitus: a randomized, placebo-controlled study. Clin Ther [39] Aubin HJ, Farley A, Lycett D, Lahmek P, Aveyard P. Weight gain in smokers
2003;25:194–209. after quitting cigarettes: meta-analysis. BMJ 2012;345:e4439.
[24] Rosenstock J, Hassman DR, Madder RD, Brazinsky SA, Farrell J, [40] Canga N, De Irala J, Vara E, Duaso MJ, Ferrer A, Martinez-Gonzalez MA.
Khutoryansky N, et al. Repaglinide versus nateglinide monotherapy: a Intervention study for smoking cessation in diabetic patients: a randomized
randomized, multicenter study. Diabetes Care 2004;27:1265–70. controlled trial in both clinical and primary care settings. Diabetes Care
[25] Saloranta C, Hershon K, Ball M, Dickinson S, Holmes D. Efficacy and 2000;23:1455–60.
safety of nateglinide in type 2 diabetic patients with modest fasting hyper- [41] Persson LG, Hjalmarson A. Smoking cessation in patients with diabetes
glycemia. J Clin Endocrinol Metab 2002;87:4171–6. mellitus: results from a controlled study of an intervention programme
[26] Schernthaner G, Grimaldi A, Di Mario U, Drzewoski J, Kempler P, in primary healthcare in Sweden. Scand J Prim Health Care 2006;24:
Kvapil M, et al. GUIDE study: double-blind comparison of once-daily 75–80.

Downloaded for Anonymous User (n/a) at CK Mobile App Test from ClinicalKey.com by Elsevier on March 16, 2024.
For personal use only. No other uses without permission. Copyright ©2024. Elsevier Inc. All rights reserved.