The association between obesity and diabetes mellitus was recognized as early as the 1920s, 1 and numerous studies since that time have shown that the obese are at higher risk of developing diabetes in adulthood than the nonobese. 2–4 Although it has not been convincingly demonstrated that weight loss in already overweight individuals can reduce the risk of adult diabetes, weight reduction in both obese diabetics and nondiabetics has been shown to decrease blood glucose concentrations and improve insulin sensitivity. 5–7 Furthermore, the improvement in glucose response associated with weight loss seems to occur whether the weight loss is a result of caloric restriction or increased activity, although exercise seems to be the more effective means of improving insulin sensitivity. 8,9
The improvement in metabolic parameters after weight reduction suggests that sustained weight loss may reduce diabetes risk, but previous epidemiologic data are sparse and inconsistent. In a few studies, weight loss has been associated with a decreased risk of diabetes, but then only in some body mass index (BMI) subgroups. 3,10,11 In fact, little to no effect was seen among the more obese subjects in two of the three studies cited above, 10,11 and some investigators found no effect of weight loss at all. 12,13
We used follow-up data from the Framingham Study to estimate the effect of sustained long-term weight loss among overweight individuals on the risk of subsequent diabetes and to compare that with the effect of weight loss that was not sustained.
Subjects and Methods
Since 1948, the surviving members of the original 5,209 subjects in the Framingham Study have been examined every 2 years. The clinic examinations consisted of a structured interview with a detailed medical history and risk behavior assessment, a physical examination, and laboratory and other measurements. At the initial examination, basic demographic information such as education and occupation was also obtained. For these analyses, we include follow-up data through the 20th biennial examination (1988–1989), during which time outcome ascertainment is complete and validated.
There were 1,574 men and women age 30–50 years with a BMI ≥27. Of these, 1,267 survived and had weight measurements taken over 16 years. We excluded an additional 31 subjects with missing covariate information, 65 with prevalent diabetes mellitus (at baseline or occurring during the initial 16-year weight-change period), and 6 who were lost to follow-up. Of the remaining 1,165 men and women, we selected 618 (333 men and 285 women) who either lost weight or had stable weight over 16 years.
At every biennial examination, weight was measured without shoes using a standard balance scale; height was measured at several examination visits. BMI for each examination visit was calculated as measured weight (kg) per average adult height of the subject (m2) before age 60.
The diagnosis of incident diabetes mellitus was made on the basis of one of the following: a history of diabetes mellitus diagnosed by a physician during follow-up, treatment with insulin or oral hypoglycemic agents, or a casual blood glucose level of 200 mg/100 ml (11.1 mmol/L) or more. All self-reported diagnoses of diabetes were confirmed by review of hospital or physician records.
We examined the following confounding variables: age, sex, education, baseline BMI, and height, as well as mean physical activity score, mean number of cigarettes smoked per day, and mean alcohol intake (ounces of pure alcohol per week) during the initial 8-year exposure period. The physical activity index was calculated as the number of self-reported hours per day spent doing moderate and vigorous activities multiplied by a numeric weight derived from the oxygen consumption required (liters per minute) for that activity. 14 Cigarette smoking and alcohol intake were assessed by self-report.
For this analysis, the first examination at which subjects were 30–50 years old and had a BMI ≥27 was considered their baseline examination. For most subjects, this was examination 1 in 1948. To assess the potential benefits of long-term weight loss, both sustained and nonsustained, among overweight, middle-aged adults, we first calculated weight change over 8 years beginning at the subject’s baseline examination. We included all of those with at least three of the five possible weight measures during the 8-year period. We used a simple linear regression model with all-available weight measures regressed on age to calculate each subject’s 8-year weight slope.
For those who lost more than an estimated 8 lb during the first 8 years of follow-up (>1 lb per year), we then calculated weight change over the next 8 years. The second 8-year weight-change interval began at the last examination visit of the first weight-change period. In this way, we classified each subject who initially lost weight into one of the following categories: lost/gained, lost/stable, or lost/lost. Subjects in the lost/gained category were those who did not sustain their weight loss, and those subjects in the lost/stable and lost/lost categories were considered to have sustained weight loss. Subjects with stable weight during both 8-year periods served as the referent group for all analyses.
Follow-up for diabetes mellitus events began at the last examination in the second weight-change period and continued until one of the following censoring events: incident diabetes mellitus, death, loss to follow-up, or examination 20 (the final examination cycle for which validated data were available). Diabetes mellitus incidence rates were calculated as the number of incident cases that occurred during the follow-up period divided by the amount of person-time in each of the four weight-change categories (stable/stable, lost/gained, lost/stable, and lost/lost).
We compared the rates of diabetes mellitus for subjects in each weight-loss category with those in the referent category (stable/stable) using Cox proportional hazards models to estimate crude and adjusted relative risk (RR) and 95% confidence interval (CI). 15 Finally, we used these same analytic methods to examine the effect of the total amount of weight lost during the initial weight-change period on diabetes risk.
The characteristics of subjects in each of the four weight-change categories are shown in Table 1. Subjects with stable weight or who lost weight and kept it off (lost/stable) were more likely to be male, and those whose regained the lost weight (lost/gained) were more frequently female. Those with stable weight were also less likely to smoke and less likely to be sedentary.
Table 2 provides the baseline BMI, diabetes incidence rates, and crude and adjusted relative risk estimates for each category of weight change. In this table, we show three different BMI cutpoints for characterizing obesity: BMI ≥27, ≥28, and ≥29. Regardless of the cutpoint used to define obesity, those with stable weight had a lower baseline BMI than those who lost weight. For those with a BMI ≥27 and stable weight, the incidence of diabetes mellitus was 8.1 per 1,000 person-years. Men and women at the same level of obesity with sustained weight loss had a 37% lower risk of diabetes (RR = 0.63; 95% CI = 0.34–1.2) than those with stable weight. Those who lost weight but then regained it had about a 30% higher risk of diabetes than those with stable weight (RR = 1.3; 95% CI = 0.70–2.4).
We also examined the relative risk estimates for the two separate categories of subjects with sustained weight loss: those who lost weight in the first interval and kept it off in the second (lost/stable) and those who lost weight in both intervals (lost/lost). Although the number of cases is sparse in these weight-loss categories, men and women who lost weight in both weight-change periods (lost/lost) experienced a lower risk of diabetes than those who lost weight only in the first interval (lost/stable).
As shown in Table 2, the protective effect of weight loss increases with increasing BMI. For example, sustained weight loss led to a 37%, 55%, and 62% reduction in diabetes risk for a BMI ≥27, ≥28, and ≥29, respectively. Finally, especially among the more obese subjects (BMI ≥29), there is no indication of an adverse effect of weight loss that was not sustained (lost/gained).
In Table 3, we examine the effects of the amount of weight lost during the first 8 years, again using those whose weight was stable over the entire 16-year period as the referent category. Among subjects whose weight loss was not sustained, there was no reduction in diabetes risk. Among those whose weight loss was sustained, a loss of 8.1–15 lb (17.8–33 kg) was associated with a 33% lower risk of diabetes, and a loss of more than 15 lb (33 kg) was associated with a 51% lower risk of diabetes compared with those whose weight was stable.
This study demonstrates that sustained weight loss reduces the risk of diabetes mellitus in obese middle-aged adults, and this effect increases with increasing level of obesity. Subjects who lost more than 1 lb per year during the initial 8-year period and kept the weight off during the second period and those who continued to lose weight during the second 8-year exposure period had a substantially lower long-term risk of developing diabetes than did those with stable weight. Weight loss that was not sustained, regardless of the amount of weight lost, had little effect on diabetes risk. On the other hand, if weight loss was sustained, higher levels of weight loss were more strongly protective against diabetes mellitus than were lower levels.
The protective effect of weight loss among obese subjects that we saw in this study conflicts with the results of some previous studies. Although the Nurses’ Health Study found an overall beneficial effect of weight loss from late adolescence to middle adulthood, there was little protective effect seen for the most obese women. 10 In the First National Health and Nutrition Examination Survey (NHANES I) cohort, weight loss over approximately 10 years among those with a BMI ≥29 at baseline was not associated with a reduced risk of diabetes mellitus over the next 10 years. 11
There are a number of important differences between earlier studies and the current analyses, including different ages of subjects at baseline, varying follow-up intervals, use of self-reported measures of weight, and different numbers of weight measures for the estimation of long-term weight change. Previous studies that had only two measures of weight could not separate sustained from nonsustained weight loss and, in addition, may suffer from random error in the classification of weight change, factors that may explain some of the earlier null results. 3,10–13 In the Framingham Study, weight was measured at 2-year intervals, and long-term weight change was estimated using a separate linear regression model to obtain each subject’s slope of weight change during two consecutive periods. This approach should reduce random error and facilitate the detection and description of true weight change. It will also lessen the impact of isolated fluctuations in weight occurring at the beginning or end of the interval, without ignoring the impact of weight fluctuations during the interval.
Because weight loss is frequently not sustained and many questions remain concerning the possible detrimental effects of regaining lost weight, we felt that it was important to address this question. In this group of obese men and women, we found no convincing evidence to suggest that unsuccessful attempts at weight loss are more detrimental than stable weight with regard to the development of diabetes mellitus.
To study the potential beneficial effect of weight loss, we would ideally like to separate voluntary from involuntary weight loss, as the latter may have resulted from illness. Earlier analyses from the Framingham Study have demonstrated that those who lost weight had a higher prevalence of cardiovascular disease, cancer, diabetes mellitus, and other diseases than those who had stable weight or who gained weight. 16 Although these earlier analyses have shown a beneficial effect of weight loss on blood pressure and serum total cholesterol, its effect on incident disease has not been examined. Although weight loss of any origin may reduce diabetes mellitus risk through the impact of reduced fat mass on glucose metabolism, the beneficial effects of weight loss may originate from direct positive effects of activity or diet on metabolic pathways. To reduce the possibility that weight loss was a consequence of subclinical diabetes or some other illness, we repeated all analyses excluding the first 2 years of follow-up after weight change. These results were unchanged, suggesting that subclinical disease had no effect on our earlier analyses. In any event, it is unlikely that subclinical disease could explain a protective effect of weight loss, because the subclinical illness per se is not likely to protect against development of diabetes mellitus. Therefore, any confounding that might be present from existing illness is likely to have led to an underestimation of the true protective effect of weight loss. Nevertheless, we are not able to examine directly the specific causes of weight loss in this study and cannot address whether weight loss associated with increased activity has the same effect as weight loss resulting from a restriction in energy intake. Furthermore, we have no information on how particular dietary patterns might impact weight loss or diabetes risk.
The National Institutes of Health recently released a report calling for a randomized clinical trial to estimate the long-term health benefits of intentional weight loss as well as the specific effects of sustained weight loss on morbidity and mortality. 17 The current analyses from the Framingham Study provide substantial new evidence that sustained weight loss in obese, middle-aged adults protects against the development of diabetes. It further shows that the amount of weight loss required to substantially reduce the risk of diabetes is quite modest, in the range of 1 to 2 lb per year. Thus, small amounts of weight loss, sustained over a long period of time, have the potential to benefit substantial numbers of obese individuals.
We thank Halit Silbershatz for preparation of the initial Framingham Study data files and Halit Silbershatz and Wendy McKelvey for preliminary data analysis assistance and thoughtful suggestions.
1. Joslin EP. The prevention of diabetes mellitus. JAMA 1921; 76:79–84.
2. Knowler WC, Pettitt DJ, Savage PJ, Bennett PH. Diabetes incidence in Pima Indians: contributions of obesity and parental diabetes. Am J Epidemiol 1981; 113:144–156.
3. Chan JM, Rimm EB, Colditz GA, Stampfer MJ, Willett WC. Obesity, fat distribution, and weight gain as risk factors for clinical diabetes in men. Diabetes Care 1994; 17:961–969.
4. Wilson PWF, Anderson KM, Kannel WB. Epidemiology of diabetes mellitus in the elderly: the Framingham Study. Am J Med 1986; 80(suppl 5a):3–9.
5. Pi-Sunyer FX. Weight and non-insulin-dependent diabetes mellitus. Am J Clin Nutr 1996; 63suppl:426S–429S.
6. Ashley FW Jr, Kannel WB. Relation of weight change to changes in atherogenic traits: the Framingham Study. J Chron Dis 1974; 27:103–114.
7. DeFronzo RA. The triumvirate: β-cell, muscle, liver: a collusion responsible for NIDDM. Diabetes 1988; 37:667–687.
8. Dengel DR, Pratley RE, Hagberg JM, Rogus EM, Goldberg AP. Distinct effects of aerobic exercise training and weight loss on glucose homeostasis in obese sedentary men. J Appl Physiol 1996; 81:318–325.
9. Yamanouchi K, Shinozaki T, Chikada K, Nishikawa T, Ito K, Shimizu S, Osawa N, Suzuki Y, Maeno H, Kato K, Oshida Y, Sato Y. Daily walking combined with diet therapy is a useful means for obese NIDDM patients not only to reduce body weight but also to improve insulin sensitivity. Diabetes Care 1995; 18:775–778.
10. Colditz GA, Willett WC, Rotnitzky A, Manson JE. Weight gain as a risk factor for clinical diabetes mellitus in women. Ann Intern Med 1995; 122:481–486.
11. Ford ES, Williamson DF, Liu S. Weight change and diabetes incidence: findings from a national cohort of US adults. Am J Epidemiol 1997; 146:214–222.
12. Modan M, Karasik A, Halkin H, Fuchs Z, Lusky A, Shitrit A, Modan B. Effect of past and concurrent body mass index on prevalence of glucose intolerance and type 2 (non-insulin-dependent) diabetes and on insulin response: the Israel study of glucose intolerance, obesity and hypertension. Diabetologia 1986; 29:82–89.
13. Holbrook TL, Barrett-Conner E, Wingard DL. The association of lifetime weight and weight control patterns with diabetes among men and women in an adult community. Int J Obes 1989; 13:723–729.
14. Kannel WB, Sorlie P. Some health benefits of physical activity: the Framingham Study. Arch Intern Med 1979; 139:857–861.
15. Kalbfleisch JD, Prentice RL. The Statistical Analysis of Failure Time Data. New York: John Wiley and Sons, 1980; 70–118.
16. Higgins M, D’Agostino R, Kannel W, Cobb J. Benefits and adverse effects of weight loss: observations from the Framingham Study. Ann Intern Med 1993; 119 (7 pt 2):758–763.
17. Yanovski SZ, Bain RP, Williamson DF. Report of a National Institutes of Health-Centers for Disease Control and Prevention workshop on the feasibility of conducting a randomized clinical trial to estimate the long-term health effects of intentional weight loss in obese persons. Am J Clin Nutr 1999; 69:366–372.