Lifestyle factors and mortality among adults with diabetes: findings from the European Prospective Investigation into Cancer and Nutrition–Potsdam study

Authors

  • Ute NÖTHLINGS,

    1. Department of Epidemiology, German Institute for Human Nutrition (DIfE), Potsdam-Rehbrücke, Nuthetal
    2. Epidemiology Section, Institute of Experimental Medicine, Christian-Albrechts University Kiel, Kiel, Germany
    Search for more papers by this author
  • Earl S. FORD,

    1. Division of Adult and Community Health, National Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control and Prevention, Atlanta, Georgia, USA
    Search for more papers by this author
  • Janine KRÖGER,

    1. Department of Epidemiology, German Institute for Human Nutrition (DIfE), Potsdam-Rehbrücke, Nuthetal
    Search for more papers by this author
  • Heiner BOEING

    1. Department of Epidemiology, German Institute for Human Nutrition (DIfE), Potsdam-Rehbrücke, Nuthetal
    Search for more papers by this author

  • The findings and conclusions in this article are those of the authors and do not represent the official position of the Centers for Disease Control and Prevention.

Heiner Boeing, German Institute of Human Nutrition Potsdam-Rehbrücke, Department of Epidemiology, Arthur-Scheunert-Allee 114-116, 14558 Nuthetal, Germany.
Tel: +49 33200 88710
Fax: +49 33200 88721
Email: boeing@dife.de

Abstract

Background:  Healthy lifestyle behaviors are among the cornerstones of diabetes self-management, but the extent to which healthy lifestyle factors could potentially prevent premature mortality among people with diabetes remains unknown. The aim of the present study was to estimate the reduction in mortality that could be achieved if people with diabetes did not smoke, had a body mass index <30 kg/m2, performed physical activity for ≥3.5 h/week, reported better dietary habits, and consumed alcohol moderately.

Methods:  A prospective cohort study of 1263 German men and women with diabetes aged 35–65 years who were followed for an average of 7.8 years was used and multivariate Cox regression models for all-cause and cause-specific mortality were calculated.

Results:  Approximately 7% of study participants had no favorable factors, 24% had one, 35% had two, and 34% had three or more. Compared with participants who had no favorable factors, the reduction in risk was 34% [95% confidence interval (CI) 19%, 63%] for those with one favorable factor, 49% (95% CI 9%, 71%) for those with two, and 63% (95% CI 31%, 80%) for those with three or more. Furthermore, a competing risk analysis did not show any difference in the inverse associations with mortality due to cardiovascular disease, cancer, or other causes.

Conclusions:  Favorable lifestyle factors can potentially achieve substantial reductions in premature mortality among people with diabetes. Our results emphasize the importance of helping people with diabetes optimize their lifestyle behaviors.

Introduction

Diabetes is a growing health problem in the US and throughout the world.1 People with diabetes are at high risk of micro- and macrovascular complications and therefore have a reduced life expectancy.2 Maximizing life expectancy by preventing early death is a priority for diabetes management and recommendations for a healthy lifestyle similar to those given to the general public are among the cornerstones of diabetes self-management.3–5 However, studies analyzing lifestyle behaviors as risk-modifying factors for mortality or complications in diabetes are scarce.6

A number of prospective studies conducted in samples of the general population have reported that adults who optimize several lifestyle factors substantially reduce their risk for diabetes, cardiovascular disease, and all-cause mortality.7–15 However, to the best of our knowledge, this relationship has not been demonstrated among individuals with diabetes. Therefore, the aim of the present study was to examine the potential for risk reduction of all-cause and cause-specific mortality associated with favorable lifestyles from an observational study.

Methods

In Potsdam, Germany, men aged 40–65 years and women aged 35–65 years from the general population were invited to join the European Prospective Investigation into Cancer and Nutrition (EPIC)–Potsdam study.16,17 Between 1994 and 1998, 27 548 adults consented to participate. The baseline examination included anthropometric and blood pressure measurements performed by trained personnel, a personal interview that included questions on prevalent diseases, a questionnaire on sociodemographic and lifestyle characteristics, and a food frequency questionnaire. Participants were followed actively (i.e. they were contacted biennially to update information on incident diseases and selected lifestyle factors). The response rates for each of the four waves of follow-up exceeded 90%. Mortality was ascertained by inquiries to municipality registries, regional health departments, physicians, or hospitals. Consent was obtained from all those participating in the study and the study was approved by the Ethics Committee of the State of Brandenburg, Germany.

Self-reported diagnoses of diabetes were confirmed by using additional sources of information, such as the use of antidiabetic medication, consistent self-report in follow-up, or reconfirmation following additional contact with the participant. A total of 1263 participants (708 men and 555 women) were included in the analysis after excluding participants <35 years or >65 years at recruitment (target age range for the EPIC study), those who never completed a follow-up questionnaire (we were not able to calculate a censoring date, because we did not consider the assumption that these participants were still alive reliable), and those who had missing information for a lifestyle behavior or covariate and whose diabetes diagnosis could not be verified.

The study investigated five factors [smoking status, body mass index (BMI), physical activity, diet, and moderate alcohol consumption]. Each behavior was dichotomized into two categories: favorable (1 point) and unfavorable (0 point). Definitions of favorable lifestyle factors were as follows: (i) never smoking; (ii) a BMI <30 kg/m2 (based on measured weight and height); (iii) physical activity ≥3.5 h/week; (iv) a value above the median of the sum of the z-scores of the consumption of fruits and vegetables and whole-grain bread, minus the z-score of red meat intake (processed and unprocessed) reported on a food frequency questionnaire; and (v) moderate alcohol intake at recruitment (5–15 g/day for women and 5–25 g/day for men).

We included the following covariates that were assessed with a self-administered questionnaire and a personal interview during the baseline data collection: age, sex, educational attainment (no vocational training completed, vocational training completed, technical school, university), occupational status (unemployed, employed full-time, employed part-time, work by the hour, retired), prevalent myocardial infarction, prevalent stroke, prevalent hypertension (based on either self-report, the use of characteristic medications, or the actual blood pressure measured at recruitment), age at the time of diagnosis of diabetes, duration of diabetes, and use of insulin or oral antidiabetic medication. In addition, we investigated the role of actual blood pressure by adding to the model systolic blood pressure (SBP), which was measured at recruitment. For the latter models, data were available for 123 deceased individuals and 1173 individuals who were still alive.

Cox proportional hazards analysis was used to estimate hazard ratios (HR) and 95% confidence intervals (CI). Analyses were stratified on age at baseline and adjusted for the covariates listed above. Competing risk models were computed to compare the impact of adhering to favorable lifestyle factors on premature death due to either cardiovascular diseases (CVD; 10th Revision of the International Classification of Diseases18 (ICD10) codes I00–I99; n = 48), cancer (ICD10 codes C00–D49; n = 48), or other known causes (= 35).19 Interaction terms with sex, use of insulin as an indicator of Type 1 diabetes, or SBP were analyzed. Because no fasting blood sample was available, we were unable to analyze glycemia in more detail. We calculated rate advancement periods for the number of favorable lifestyle factors.20 All analyses were performed using sas version 9.1 (SAS Institute, Cary, NC, USA).

Results

In our analytic sample of 1263 people with diabetes followed for an average of 7.7 years, mean (±SD) age was 57 ± 6 years. The distribution of the number of favorable lifestyles was as follows: 7% had zero factors, 24% had one, 35% had two, and 34% had three or more (Table 1).

Table 1.   Baseline characteristics of the 1263 participants with diabetes, European Prospective Investigation into Cancer and Nutrition–Potsdam Study
 No. favorable lifestyle factors from never smoking, good diet, BMI <30 kg/m2, ≥3.5 h/week physical activity, and lifetime moderate alcohol consumption
012≥3
  1. Where appropriate, data are shown as the mean ± SD.

  2. *Based on self-report, self-reported medication, or actual blood pressure measured at study entry.

  3. Systolic blood pressure was measured at baseline and was available for 121 cases and 1072 non-cases.

  4. BMI, body mass index. Cases, refers to deceased individuals. Non-cases, refers to individuals who are still alive.

n 72267399395
No. events 19 36 46 29
% Men 70 63 55 49
Age at baseline (years) 56 ± 7 57 ± 6 58 ± 6 57 ± 7
Age at diagnosis of diabetes (years) 50 ± 9 50 ± 9 50 ± 9 50 ± 10
Duration of diabetes (years)  7.2 ± 7.8  7.8 ± 7.6  7.9 ± 7.5  8.0 ± 7.5
Duration of follow-up (years)  7.5 ± 2.0  7.7 ± 2.1  7.6 ± 2.1  8.0 ± 1.7
BMI (kg/m2) 34.1 ± 3.3 30.6 ± 4.8 29.6 ± 5.3 27.4 ± 3.9
% Using insulin 18 16 16 17
% Using other diabetes-related medications 52 48 42 40
% With prevalent myocardial infarction 19  8  8  7
% With prevalent stroke  4  3  7  2
% With prevalent hypertension* 91 81 80 76
SBP (mmHg)146 ± 21143 ± 19141 ± 19138 ± 18
Education (%)
 No vocational training completed 47 51 45 46
 Vocational training 21 23 25 21
 Technical school/university 32 27 31 32
Occupational status (%)
 Unemployed 29 35 28 34
 Employed full-time  2  4  4  4
 Employed part-time  1  4  5  3
 Hourly worker 45 48 51 47
 Retired 23 10 12 12

Compared with participants with zero index points, the risk of premature death decreased progressively as the number of favorable factors increased (Table 2). Thus, adults with three or more such factors had a 63% (95% CI 31%, 80%) reduced risk for all-cause mortality (P < 0.001 for linear trend). After adjustment for actual measured blood pressure, adoption of three or more favorable lifestyle factors was still inversely associated with mortality [HR 0.42, 95% CI 0.22–0.80; P = 0.002 for trend]. No interaction with sex, insulin therapy, or SBP was present. The reduced risk for those with three or more favorable factors corresponded to approximately 3 years of life.

Table 2.   Hazard ratios (95% confidence intervals) for lifestyle habits and mortality among 1263 participants with diabetes, European Prospective Investigation into Cancer and Nutrition–Potsdam Study
ExposuresNo. casesPYSex-adjusted model*Model IModel II
  1. *All models are stratified on age and adjusted for sex.

  2. Additionally adjusted for educational status (no vocational training completed, vocational training completed, technical school, university), occupational status (unemployed, employed full-time, employed part-time, hourly worker, retired), and prevalent myocardial infarction, stroke, or hypertension.

  3. Adjusted for all factors in Model I and the duration of diabetes, age at time of diagnosis of diabetes, the use of insulin, and the use of oral antidiabetic medication.

  4. §Lifestyle factors include never smoking, good diet, body mass index <30 kg/m2, ≥3.5 h/week physical activity, and moderate alcohol consumption.

  5. **See Methods for details.

  6. ††P-values in the sex-adjusted model, Model I and Model II for differences between hazard ratios (HR) for cardiovascular disease (CVD) and other causes of death are 0.38, 0.57, and 0.46, respectively; between HR for cancer and other causes of death 0.87, 0.96, and 0.97, respectively; and between HR for CVD and cancer 0.32, 0.53, and 0.48, respectively.

  7. ICD10, 10th Revision of the International Classification of Diseases;18 PY, person-years.

All-cause mortality
 No. favorable lifestyle factors§
  0196841 (reference)1 (reference)1 (reference)
  13623240.58 (0.33–1.03)0.64 (0.36–1.15)0.66 (0.37–1.19)
  24633860.47 (0.27–0.81)0.50 (0.28–0.88)0.51 (0.29–0.91)
  ≥32933850.32 (0.18–0.58)0.36 (0.20–0.67)0.37 (0.20–0.69)
 Smoking
  Ever1025792111
  Never2839870.41 (0.26–0.65)0.42 (0.27–0.66)0.42 (0.27–0.67)
 BMI
  ≥30 kg/m2603958111
  <30 kg/m27058220.71 (0.50–1.00)0.80 (0.56–1.15)0.81 (0.56–1.17)
 Diet**
  Unfavorable754937111
  Favorable5548430.78 (0.54–1.12)0.77 (0.54–1.12)0.75 (0.52–1.08)
 Physical activity     
  Inactive1007369111
  Active3024100.97 (0.64–1.46)0.95 (0.63–1.44)1.00 (0.66–1.53)
 Alcohol consumption (moderate vs other)
  Other92647811 
  Moderate3833010.69 (0.47–1.03)0.72 (0.49–1.08)0.73 (0.49–1.08)
Cause-specific mortality††
 CVD (ICD10 codes I00–I99)
  One more factor48 0.64 (0.47–0.87)0.69 (0.51–0.94)0.67 (0.49–0.92)
 Cancer (ICD10 codes C00–D49)
  One more factor47 0.79 (0.61–1.02)0.79 (0.60–1.03)0.79 (0.59–1.04)
 Other causes
  One more factor34 0.77 (0.60–0.98)0.78 (0.59–1.04)0.79 (0.59–1.06)

Never smoking was the only significant predictor among the five factors (adjusted HR = 0.42; 95% CI 0.27, 0.67; Table 2). Physical activity alone was the weakest predictor for all-cause mortality when using the cut-off points established in the present study. Omission of physical activity in our index showed the following HR (95% CI) for zero, one, two, and three or more favorable factors: 1 (reference); 0.58 (0.33–1.00); 0.46 (0.26–0.79); and 0.29 (0.15–0.56), respectively.

No significant differences in HR for an increase of one more favorable behavior across different causes of death (CVD, cancer, other) were observed (Table 2), but HR were lowest and only significant for death due to CVD (0.67; 95% CI 0.49–0.92).

Discussion

Our results showed a progressive decrease in premature mortality as the number of favorable lifestyle factors increased among patients with diabetes, a finding that is consistent with previous results from cohort studies in the general population. Thus, our results suggest that it is important to emphasize to diabetic patients that adopting or maintaining healthy lifestyle behaviors is an important component of diabetes self-management. Such messages are consistent with positions adopted by the American Diabetic Association for smoking, physical activity, and dietary practices.4

Studies focusing on single lifestyle factors and mortality in individuals with diabetes have shown inverse associations with physical activity,21–23 and fruit and vegetables intake,19 moderate alcohol consumption24 and positive associations with smoking.25 A moderate inverse effect of physical activity on blood glucose control was found in a meta-analysis of interventions studies.26 However, studies so far are scarce and the combination of lifestyle factors has, to our knowledge, not been addressed.

Many diabetic adults in the US still smoke, are obese, are inadequately active, and eat insufficient fruits and vegetables and too little dietary fiber.15,27–31 This has also been seen in the EPIC–Potsdam study32,33 and can be expected to be similar for other European countries. Thus, a great deal of progress in optimizing lifestyle factors in diabetic adults remains to be achieved.

Management of both glycemia and blood pressure has been shown to be important in reducing the risk of complications in diabetes.34 Lifestyle factors may also influence blood pressure; therefore, blood pressure may be an intermediate in the causal pathway in our analysis. However, we also calculated a model controlling for actual measured blood pressure that showed weakened, but still inverse, associations with lifestyle factors.

The relatively small size of the cohort limited more detailed analyses and the number of incident CVDs was too small for meaningful analysis. Furthermore, information concerning smoking, physical activity, and diet was self-reported and subject to measurement error and resulting misclassification. This could have been particularly pronounced because diabetic individuals are largely overweight and potentially underreport dietary intake.35 By dichotomizing the healthy factors and creating a simple index variable, we were unable to explore the whole continuum of risk. However, this approach has been used successfully in the past.32 The dietary variable was constructed as a relative index, not taking absolute intakes into account. However, diet was measured with a food-frequency questionnaire, which does not allow the interpretation of intakes on an absolute scale.36 Furthermore, physical activity as a single variable was not associated with mortality in our study population, questioning its inclusion on the overall index variable. However, based on published studies,22,23 we had decided a priori to construct a hypothesis-driven index that included physical activity as a favorable lifestyle factor. We were able to confirm inverse associations when we excluded physical activity from the index. Similar indices have been used previously for analyses of the EPIC-Potsdam cohort32 and other cohorts.7–15 In addition, the findings are only generalizable to the German population.

Further studies are needed to confirm our results in other diabetic cohorts. For now, our findings from an observational study emphasize that favorable lifestyle factors may exert a powerful influence in reducing premature mortality among people with diabetes and could be of considerable public health importance given the increasing prevalence of individuals living with diabetes in populations across the world.

Acknowledgments

The EPIC–Potsdam study was funded, in part, by the Germany Cancer Aid and the Federal Ministry of Education and Research. This study was supported by an European Federation for the Study of Diabetes/Sanofi-Aventis grant.

Disclosure

This manuscript has not been published previously and is not under consideration for publication elsewhere. The authors report no conflict of interest.

Ancillary