Healthy lifestyle and life expectancy free of cancer, cardiovascular disease, and type 2 diabetes: prospective cohort study

, , Comments Off on Healthy lifestyle and life expectancy free of cancer, cardiovascular disease, and type 2 diabetes: prospective cohort study

  • Study: Healthy Lifestyles Yield Benefits for Disease-Free Life ExpectancyIntroduction

    The average life expectancy in the world has increased substantially in the past few decades.1 The aging of the population has led to a high prevalence of chronic diseases such as diabetes, cardiovascular disease, and cancer. Although people live longer, older individuals often live with disabilities and chronic diseases.2 People with chronic diseases including cancer, cardiovascular disease, and diabetes have a shorter life expectancy than do their peers without these chronic conditions.345 Estimates of the loss in life years due to these chronic conditions range from 7.5 to 20 years, depending on the methods used and the characteristics of the study population.345

    Modifiable lifestyle factors including smoking, physical activity, alcohol intake, body weight, and diet quality affect both total life expectancy and incidence of chronic diseases.6789 Studies have shown that smoking, inactivity, poor diet quality, and heavy alcohol consumption contribute up to 60% of premature deaths and 7.4-17.9 years’ loss in life expectancy.7101112131415 Nevertheless, little research has looked at how a combination of multiple lifestyle factors may relate to life expectancy free from the major diseases of diabetes, cardiovascular disease, and cancer.1516171819

    Because estimates of life expectancy free of chronic diseases take into account both morbidity and mortality, these estimates can be useful metrics for health professionals and the general public, as well as enabling policy makers to better estimate future healthcare costs and to plan for healthcare needs.202122 In a previous analysis, we estimated the effect of healthy lifestyles on the overall life expectancy.23 In this study, we examine the effect of healthy lifestyle factors on life expectancy free of cancer, cardiovascular disease, and type 2 diabetes, using data from up to 34 years of follow-up in the Nurses’ Health Study (NHS) and 28 years of follow-up in the Health Professions Follow-up Study (HPFS).

  • Methods

    Study population

    This study was embedded in the NHS and the HPFS. The NHS began in 1976, when 121 700 female nurses aged 30-55 years were included and provided information on medical, lifestyle, and other health related variables.24 In 1980, 92 468 nurses also completed a validated food frequency questionnaire. The HPFS was established in 1986, when 51 529 male US health professionals (dentists, optometrists, osteopaths, podiatrists, pharmacists, and veterinarians) aged 40-75 years completed a mailed questionnaire about their medical history and lifestyle.25 In both cohorts, self administered questionnaires have been sent every two years to update the information and identify newly diagnosed cases of various diseases. For this analysis, we used 1980 as the baseline for the NHS and 1986 for the HPFS. We excluded participants already diagnosed as having any of the three outcomes (cancer, cardiovascular disease, and diabetes, n=15 118), those with implausible energy intakes (women: <500 or >3500 kcal/day; men: <800 or >4200 kcal/day), and those with missing values for body mass index, physical activity, alcohol, or smoking at baseline (n=17 317), leaving 111 562 participants (73 196 women and 38 366 men) for analysis. Participants who had missing lifestyle factors at baseline had similar baseline characteristics to those without missing information (supplementary table A).

    The NHS and HPFS cohorts were followed across the follow-up periods using similar questionnaires on diet, exercise, smoking status, and other factors (questions on the use of postmenopausal hormone replacement therapies and reproduction related questions were asked in the NHS only). Information on age, ethnicity, use of multivitamins, regular use of aspirin, postmenopausal hormone use (NHS only), and the presence or absence of a family history of diabetes, cancer, or myocardial infarction (in first degree relatives) was collected via biennial questionnaires.

    Assessment of lifestyle behaviors

    We derived a healthy lifestyle score based on information on five lifestyle factors—diet, smoking, physical activity, alcohol consumption, and body mass index (BMI). Diet was assessed in the NHS and HPFS by using a validated food frequency questionnaire assessing how often, on average, a participant had consumed a specified amount of a list of foods during the previous year.24 Quality of diet was assessed using the Alternate Healthy Eating Index (AHEI) score, which is significantly associated with risk of cardiovascular disease and other chronic diseases in the general population.26 We defined a healthy diet as an AHEI score in the top 40% of each cohort distribution.7 Physical activity levels were assessed using a validated questionnaire and updated every two to four years.27 We estimated the number of hours per week spent in moderate to vigorous activities (including brisk walking) requiring the expenditure of at least 3 metabolic equivalents of task (METs) per hour. We classified low risk as at least 30 minutes of moderate or vigorous activity daily (3.5 h/week). Height and weight were self reported and used to calculate BMI as weight (kg) divided by height (m2). We defined a healthy body weight as a BMI in the range of 18.5-24.9.

    Self reported smoking status and the number of cigarettes smoked were updated biennially. We defined never smokers as participants who reported never smoking, current smokers as participants who reported active smoking on the questionnaire, and ever smokers as participants who had smoked in the past but did not report active smoking on the time varying questionnaire. On each questionnaire, current smokers were further classified as smoking one to 14, 15 to 24, or 25 or more cigarettes per day. The food frequency questionnaire also collected alcoholic beverage consumption, including red and white wine separately (4 ounces, increasing to 5 ounces in 2006), beer (one glass, can, or bottle), and liquor (one drink or shot). We multiplied the amount of alcohol in grams per specified portion size by servings per day, determined the midpoint of the frequency category, and summed across all beverages to estimate the average alcohol consumption (g/day). We defined moderate alcohol consumption as 5-15 g/day for women and 5-30 g/day for men, consistent with the guidelines for moderate alcohol intake in the US.28

    Our previous studies using objective measurements had documented the validity of these lifestyle data. A previous validation study showed a correlation of 0.97 between self reported weight and weight measured by a technician.25 The correlations between physical activity reported in diaries and that in the questionnaire was 0.62 in women and 0.58 in men.2729 In addition, independent associations have been observed between physical activity and several biomarkers of obesity and cardiovascular disease risk such as high density lipoprotein cholesterol, leptin, and C peptide.30 Validation studies of the food frequency questionnaire with dietary records indicated an average correlation of 0.63 for all nutrients after adjustment for energy intake and variation in the seven day dietary records.24 In addition, correlations with nutrient biomarkers in our study population provided objective evidence for the validity of the food frequency questionnaire.3132 Correlation coefficients between the food frequency questionnaire and four single week diet records for ethanol were 0.90 in the NHS and 0.86 in the HPFS.33 Ethanol intake assessed through the food frequency questionnaire was also significantly correlated with serum high density lipoprotein cholesterol (r=0.33 for NHS; r=0.38 for HPFS). Another validation study among 2485 women participating in the NHS reported that smoking level was strongly associated with toenail nicotine concentration (r=0.63).34

    For each low risk lifestyle factor, the participant received a score of 1 if he or she met the criterion for low risk, and 0 otherwise. The sum of these five scores together gave a final low risk lifestyle score ranging from 0 to 5, with higher scores indicating a healthier lifestyle. As the number of cases among the group with the highest lifestyle score was small, especially mortality among participants with prevalent diseases, we combined participants with four or five low risk factors into one group.

    Because lifestyle factors may affect mortality risk over an extended period of time, to best represent long term lifestyle we applied time varying lifestyle information during follow-up, in which mortality risks were predicted by the repeated measurements of these variables during the follow-up periods. The average response rate for the lifestyle risk factors during follow-up was approximately 94% for the NHS and 90% for the HPFS. We calculated average levels of lifestyle factors by using the latest two repeated measurements for our primary analysis of diet, physical activity, and alcohol consumption.35 For non-respondents to both questionnaires, we used the last available value carried forward. For AHEI score and alcohol consumption, we calculated the average on the basis of four year repeated measurements. Smoking status was estimated on the basis of both smoking history and most recent status updated every two years and classified into five categories: never smoking, past smoking, and current smoking of 1-14, 15-24, and 25 or more cigarettes per day. To minimize reverse causality, we applied the lifelong maximum BMI by age at risk.36 For example, we applied the maximum value of BMI at age 18 and BMI in 1980 to predict mortality between 1980 and 1982 and the maximum value of BMI at age 18, BMI in 1980, and BMI in 1982 to predict mortality between 1982 and 1984, and so forth.

    Ascertainment of deaths and non-fatal chronic diseases

    In the NHS and HPFS, most of the deaths were identified through family members or the postal system in response to the follow-up questionnaires. We searched the National Death Index to identify deaths among all study participants, with a high sensitivity (97.7%) and specificity (100%).37

    Self reported diagnoses of cancer, myocardial infarction, and stroke were collected on biennial questionnaires, and participants who reported a new diagnosis were asked for permission to acquire their medical records and pathologic reports. Study physicians, blinded to exposure information, reviewed medical records to confirm the diagnosis. We included all cancers as outcomes except non-melanoma skin cancer. Non-fatal cardiovascular disease outcomes comprised non-fatal myocardial infarction and non-fatal stroke. We classified non-fatal myocardial infarctions as confirmed if the criteria of the World Health Organization were met specifically on the basis of symptoms and either electrocardiographic changes or elevated cardiac enzyme concentrations. We classified stroke cases according to the criteria of the National Survey of Stroke, which required evidence of a neurologic deficit with a sudden or rapid onset that persisted for more than 24 hours. Pathologically confirmed cerebrovascular conditions that were caused by an infection, trauma, or malignancy were not counted as outcomes. Cancer, coronary heart disease, and stroke events for which confirmatory information was obtained by interview or letter but without access to medical records were also included as outcomes. The overall proportion of cases confirmed by medical records in the NHS and HPFS were 82.9% for cancer, 74.4% for coronary heart disease, and 64.5% for stroke.

    Cases of type 2 diabetes were identified by self report and confirmed by a validated supplementary questionnaire.3839 For cases before 1998, we applied the National Diabetes Data Group criteria. We used the American Diabetes Association diagnostic criteria for confirmation from 1998 onward. The validation of self reported type 2 diabetes diagnosis in the NHS has been documented previously.3839

    Statistical analysis

    We used population based multistate life tables to calculate the differences in life expectancy and years lived with and without major chronic diseases for each lifestyle factor and the total lifestyle factor score. To assess the association between the number of low risk factors and life expectancy free of cancer, cardiovascular disease, and type 2 diabetes, we took into account three states (disease free, presence of disease, and death), and three transitions between states (from non-disease to incident disease, from non-disease to mortality among participants free of major chronic disease, and from disease diagnosis to mortality among those with disease). Subsequently, we built multistate life tables allowing for the occurrence of the transitions. As shown in supplementary figure A, all participants started from state 0 at baseline; both the period between state 0 and state 1 (before disease diagnosis) and the period between state 0 and state 2 (without disease) contributed to the estimated life expectancy free of cancer, cardiovascular disease, or type 2 diabetes; whereas only the patients with cancer, cardiovascular disease, or type 2 diabetes after diagnosis (from state 1 to state 2) contributed to the estimated life expectancy in the presence of the major chronic disease.

    We built a total of four multistate life tables—one for a combination of cancer, cardiovascular disease, and type 2 diabetes and three for individual diseases. For each multistate life table, we first derived overall transition rates by a single year of age in the NHS and HPFS, separately, irrespective of the lifestyle factors, for three transitions4: mortality among participants free of the diseases, incident diseases, and mortality among those with the diseases. We considered only the first entry into a state and no subsequent disease event, and no reversal of state was allowed. Secondly, we calculated hazard ratios to assess the relation between the number of low risk factors and the three transitions by using Cox proportional hazards analyses. We used separate time varying Cox proportional hazards models to assess the risk of mortality without disease, risk of incident disease, and risk of mortality among participants with cancer, cardiovascular disease, and type 2 diabetes. Thirdly, we calculated proportions of low risk factors among the sub-study population for each transition. Lastly, we used the hazard ratios combined with the overall transition rates and low risk factor proportions in the multistate life table to calculate total life expectancy and life expectancy with and without diseases for each group of low risk factors. The multistate life tables started at age 50 years and closed at age 105 years. Models were adjusted for age, ethnicity, current multivitamin use, current aspirin use, status with regard to a family history of diabetes, myocardial infarction, or cancer, and, for women, menopausal status and hormone use.

    Considering potential bias resulting from changes in diet after the diagnosis of certain diseases, we did a sensitivity analysis in which we stopped updating lifestyle factors at the beginning of the interval in which the participant was diagnosed as having cancer, cardiovascular disease, or diabetes. In another sensitivity analysis, we further classified the past smokers according to the years since smoking cessation.

    We used SAS version 9.3 to analyze the data. Statistical significance was set at a two tailed P value of less than 0.05. We used Monte Carlo simulation (parametric bootstrapping) with 10 000 runs to calculate the confidence intervals of the life expectancy estimation with @RISK 7.5.

    Patient and public involvement

    No patients were involved in setting the research question or the outcome measures, nor were they involved in the design and implementation of the study. We plan to disseminate these findings to participants in our annual newsletter and to the general public in a press release.