To examine the association between body weight measures across the lifecourse and the risk of adult-onset diabetes.
We analysed data from the Glasgow Alumni Cohort and the British Women's Heart and Health Study (BWHHS). The former included 5571 men and women who had height and weight measured at university, and reported birthweight, mid- and later-life weight in a postal questionnaire. The BWHHS analysis included 4280 women who had height and weight measured in later adulthood and recalled their birthweight and early adult height and weight. Adult-onset diabetes was defined as doctor-diagnosed disease after age 30, either self-reported or abstracted from medical records.
Thirty nine women and 209 men (Glasgow Alumni study) and 314 women (BWHHS) had diabetes. Those with diabetes had lower mean birthweight than those without, although the differences were small. Individuals with diabetes were also shorter and heavier at all ages than those without diabetes. Being overweight during at least one time period in adult life was associated with an increased risk of diabetes, compared to those who were never overweight. While there was no age at which being overweight was particularly detrimental, the risk associated with being overweight was cumulative across the lifecourse.
Being overweight at any point during life is associated with an increased risk of adult-onset diabetes. The cumulative nature of this association reinforces the need to prevent the development of excess weight at an early age to reduce diabetes prevalence in coming decades.
There is considerable evidence showing an inverse association between birthweight and insulin resistance and type 2 diabetes.1, 2, 3 Being overweight or obese in adult life also increases the risk of these conditions4, 5 and there is accumulating evidence indicating that weight in pre-adult and early adult life also affect later adult-onset conditions.6, 7 However, among participants in the US Nurses Health Study body mass index, based on self-report of weight and height in a questionnaire when the participants were aged 30–55, was strongly associated with diabetes risk during eight years of follow-up, but body mass index when the nurses were 18 years of age (based on retrospective report of their earlier weight by the nurses at the baseline examination) was only weakly associated with future risk of diabetes, with this risk attenuating to the null with adjustment for later body mass index.8 Taken together, these observations imply that people who subsequently develop diabetes have a different growth trajectory or excess weight gain over the lifecourse in comparison with the rest of the population.3, 9, 10
Our aim was to describe patterns of weight over the lifecourse among individuals with and without diabetes using data from two different studies. The advantage of using data from two cohorts lies in their different designs, in terms of participants and available exposure and outcome measures.
Participants and methods
The Glasgow Alumni Study is a cohort study which has followed up University of Glasgow students who attended an annual medical examination at the Student Health Service between 1948 and 1968. The median age at university was 19 years (inter quartile range (IQR) 18–21 years). The cohort was traced through the NHS Central Register in 1998, and contacted by postal questionnaire in 2001 (10 951, 71.5%) The response rate to the questionnaire was 52.1% (n=5,612). Baseline data and full details of the study methods have been reported previously.11 The median age of participants who responded to the questionnaire was 64 years (IQR 57–70 years).
The British Women's Heart and Health Study (BWHHS) is a cohort study based on a sample of women aged 60–79 years randomly selected from general practitioner lists in 23 British towns. A total of 4286 women (60% of those invited) were interviewed between April 1999 and March 2001. Follow-up of these women has been through primary care medical record review and a follow-up questionnaire, to which 90% of survivors responded, three years after the baseline interview. Full details of participants and measurements have previously been reported.12, 13 Ethics committee approval was received and informed consent obtained for both studies.
In the Glasgow Alumni study, students’ heights and weights were measured by a physician; we used these to calculated BMI in young adulthood (median age 19 years). In the 2001 questionnaire, participants reported their birthweight and their weight at ages 20 and 40 years and current weight, from which BMI at these ages was calculated. Participants were provided with a paper tape measure and appropriate instructions for waist and hip circumference measurements, from which waist:hip ratio (WHR) was calculated.
In BWHHS, women aged 60–79 years had measures of height (to the nearest millimetre) and weight (in light clothing and without shoes, to the nearest 0.1 kg) made. Waist and hip circumferences were each measured twice, recorded to the nearest millimetre, using a flexible metal tape and the mean of the two readings was used in all analyses. BMI at age 21 is calculated from the woman's self-report of her weight at that age and her height measured at age 60–79 years. In the baseline questionnaire women were also asked to report their weight and dress size when they were aged 21.
In both the baseline questionnaire and the recent follow-up questionnaire in BWHHS, women were asked to report their birthweight. For those women who reported this on both occasions (n=690), the two measures were highly correlated (r=0.97). In the follow-up questionnaire, women who did not know their birthweight were asked to indicate into which of five birthweight categories (<5lb 8oz, 5lb 8oz to 6lb 15oz; 7lb to 8lb 15oz; 9lb to 10lb 15oz; ⩾11lb) they believed their birthweight fell. A composite birthweight variable, generated using a prioritised replacement of (i) baseline self-reported birthweight, (ii) follow-up self-reported birthweight or (iii) the mid-point of the categorical values, resulted in 2431 (57%) women having birthweight data of whom 853 provided only categorical data.
Potential confounding variables
Childhood social class, derived from reported father's occupation, was classified as I (professional) to V (unskilled manual). An additional category of ‘unemployed’ was included in BWHHS analyses. In the Glasgow Alumni cohort, adult socio-economic position was not controlled for, as it is relatively homogenous amongst this population all of whom attended university.14 Among questionnaire respondents, 80% had jobs which were classified as ‘professional’. Adult social class in BWHHS was based on own or husband's longest held occupation and classified in the same way as childhood social class.
In the Glasgow Alumni cohort, smoking was controlled for using a variable ‘ever smoked’, comprising those people who reported smoking while at university or who reported ever having smoked in the postal questionnaire. In the BWHHS, smoking data, obtained from the self-completed questionnaire, were analysed as: never, past and current (1–9, 10–19, 20–29, ⩾30 cigarettes per day).
In both studies, participants were asked whether they had ever been told by a doctor that they had diabetes, and if so, in what year they had first been diagnosed. This was in a postal questionnaire in the Glasgow cohort, and during a nurse-administered face-to-face interview in BWHHS. Those who answered yes, as well as participants who reported using insulin or oral hypoglycaemics were classified as having diabetes. In the BWHHS cases were also confirmed and additional cases identified through review of primary care medical records, which also contain all correspondence with secondary care services. There was good agreement (kappa=0.89, P<0.0001) between medical record and self-report of diabetes diagnosis. For both studies, adult-onset diabetes was defined as diabetes diagnosed in participants at age 30 or older.
We calculated mean adjusted values of weight and height at different points in the lifecourse among people who did and did not develop adult-onset diabetes. Weights and heights were adjusted for the potential confounding variables detailed above and for age (categorical variable in five-year age bands). Both the age at which the measurements were made and the age at which the outcome data were ascertained (date of questionnaire completion for the Glasgow Alumni study and date of diagnosis in the medical record, or, if this was not available, the date at which she was first diagnosed, provided by the woman in the questionnaire for BWHHS) were included. Adjusted means were estimated from linear prediction following regression of diabetes status on the body weight variable, keeping the potential confounding variables at their mean values.
Logistic regression models were used to estimate relative odds of diabetes according to differing patterns of weight, BMI and WHR. Likelihood ratio tests were used to test for supra-multiplicative statistical interaction. All models included the potential confounding variables listed above. To investigate weight patterns and change during adulthood, we created a composite score of ‘lifetime overweight’, based on BMI at each of the time points available. Overweight was defined as BMI ⩾25 kg/m2.
In the Glasgow Alumni study, 203 men and 39 women reported having diabetes and a further six men reported taking medication for diabetes. Those who did not answer the relevant questions (n=111) were considered as not having diabetes. Eighteen people with diabetes diagnosed before the age of 30 years and 23 people who did not report year of diagnosis were excluded from all analyses. The prevalence of adult-onset diabetes was therefore 3.7% (207 out of 5571 respondents). Of the 207 people with self-reported adult-onset diabetes, the median age at diagnosis was 59 years (range 30–82). Forty three of those with adult-onset diabetes (21%) reported taking insulin.
Among the 4286 women who participated in the BWHHS 314 (7.3%) were identified as having diabetes at either baseline or follow-up assessment. Six women had diabetes diagnosed before age 30 and were excluded from further analysis, yielding a prevalence of adult-onset diabetes in the BWHHS of 7.2% (308 cases out of 4280 women). The median age of diagnosis of women with adult-onset diabetes was 66 years (range 32–81). The median age at baseline assessment of those without diabetes was 69 years (range 60–79) and at follow-up assessment was 71 (range 63–82).
Mean body size measures according to diabetes outcome are shown in Table 1.
Birthweight was higher amongst people who did not have diabetes compared to those who did, although the magnitude of the differences were small (120 g in Glasgow Alumni and 70 g in BWHHS) and these differences may have arisen by chance (P=0.17 in both cohorts). In the Glasgow Alumni study, people who developed diabetes were heavier and had a higher BMI when aged 18–21 (based on measured height and weight) than those who did not. Conversely, in analyses based on recalled data, there was no indication of a difference in weight or BMI at age 20. In the BWHHS recalled weight, BMI and dress size at age 21 were all positively related to risk of adult-onset diabetes. Glasgow Alumni cohort participants who developed adult-onset diabetes reported having a higher weight at age 40. The same association existed for BMI at this age, derived using the height measure from university. No measures of weight were available in the BWHHS at this age. In the Glasgow Alumni study, self-reported current measures of weight and BMI at a mean age of 64 years were all positively related to having adult-onset diabetes, as were self-measured waist circumference and WHR. These measurements, made by a clinic nurse in BWHHS, were similarly related to diabetes, although the magnitude of the differences was greater than for the Glasgow Alumni study.
Acknowledging the small number of women with adult-onset diabetes in the Glasgow Alumni cohort (n=32), there was no strong statistical evidence of a supra-multiplicative interaction between sex and any of the measures of body size in the logistic regression model (all P-values >0.10), with the exception of recalled BMI at age 40. At that age, the difference in mean BMI between men who subsequently developed adult-onset diabetes and those who did not, adjusted for age, childhood social class and smoking was 2.0 kg/m2, whereas the difference in women was 4.1 kg/m2 (P for interaction=0.002).
One of our a priori hypotheses was that the highest risk of diabetes would be seen in people with lower birthweight who subsequently became overweight. In neither cohort was there statistical evidence of interaction between birthweight under 3 kg and adulthood overweight in determining diabetes risk. Inspection of the stratum-specific ORs for diabetes from the Glasgow Alumni study suggested that a higher risk of diabetes was present in people who were born of lower birthweight and were overweight in mid- or later adulthood. For example, comparing people born under 3 kg with those born at 3 kg or over, the OR for diabetes in those who were overweight at age 40 years was 1.54 (0.74–3.23), whereas in those people who were not overweight at age 40, it was 1.20 (0.62–2.32), P (interaction): 0.56. When people were stratified by overweight in later adulthood, the corresponding OR were 1.55 (0.83–2.32) and 1.10 (0.50–2.43), P (interaction): 0.13. However, the reverse was true in BWHHS. In those people who were not overweight at age 40, the OR for diabetes was 1.57 (0.66–3.72) and in those who were overweight it was 1.21 (0.85–1.72), P (interaction): 0.61. It appears that these differences are therefore likely to be due to chance, and that there is no modification of the effect of birthweight on diabetes risk by adulthood BMI in these two cohorts.
Changes in weight patterns according to diabetes status are shown in Table 2. Being overweight in young, mid- and later adulthood (Glasgow Alumni), or in young and later adulthood (BWHHS) was associated with a three and a half to four-fold increased risk of diabetes. Being overweight at two out of the three periods of adult life was associated with a higher risk than being overweight at one of three periods, although there was no particular age period at which being overweight appeared to confer greater risk than overweight at other ages. Data from both cohorts indicated that being overweight in young adulthood was associated with an increased risk, even amongst those people who lost weight during their mid- or later adulthood. We had no information on whether this weight loss was voluntary or involuntary.
The analyses presented here add two components to our understanding of the relationships between lifecourse weight patterns and the risk of adult-onset diabetes. First, we have demonstrated a higher risk of diabetes amongst people who were overweight in young adulthood. In these cohorts this increased risk persists even in those who are normal weight in later adulthood, compared to individuals who were normal weight both in young and later adulthood. Secondly, we have also shown a cumulative effect of being overweight throughout adult life on the risk of adult-onset diabetes. The results suggest that losing weight may not reduce disease risk to the same degree as never having been overweight.
Although the results from the two cohorts which we present are broadly consistent, in the Glasgow Alumni study there was no association between self-reported recalled weight at age 20 years and diabetes whereas for BWHHS women, self reported weight and dress size at age 21 did show substantive effects. This is despite moderate agreement between quintiles of the weight distribution measured at university and weight at age 20 recalled four decades later in the Glasgow Alumni study (kappa=0.41, P<0.001). This may be explained by more accurate recall of weight by women compared to men (kappa 0.42 vs 0.33). These results suggest that, particularly among men, studies based on recalled weight data from many decades previously will not necessarily provide accurate disease risk estimates.
The participants from the Glasgow Alumni study included in these analyses were a small proportion (17%) of the original student population. At each stage of exclusion, we have attempted to identify factors which predicted participation. Those included are healthier than the original population, since they include only survivors to a median age of 64, and are less likely to have been smokers at university. However, the broadly consistent results between the two cohorts suggest that selection bias is unlikely to explain the results we report. Ascertainment bias, i.e. the selective diagnosis of diabetes among participants who are overweight, is plausible, but would not explain the increased risks seen in people who were heavy in young adulthood and subsequently lost weight.
An important limitation of our study is the need to rely on self-reported diabetes diagnosis as the main outcome measure. As noted above, in BWHHS cases were also obtained from detailed medical record reviews conducted for all participants and there was good agreement between medical record and self-report of diabetes diagnosis. Although such data were not available in the Glasgow cohort, there is no reason to believe that in a cohort of university-educated men and women, 22% of whom attended medical school, self-report of doctor diagnosed diabetes would be less reliable that in a survey of older women. Perhaps more importantly, we had to exclude the 23 people from the Glasgow cohort who did not report the year in which they were diagnosed with diabetes, since we defined our outcome as adult-onset diabetes. It seems likely that this lack of information was probably missing at random (i.e. not related to BMI or the severity of the diabetes). This under-ascertainment will not result in a biased relative effect measure such as the OR.
The lack of strong evidence of an association between birthweight and diabetes may be due to the random misclassification arising from the recalled birthweight measures, thereby reducing the power of the study to detect differences in birthweight between participants with and without diabetes. The magnitude of the differences which we report are similar to those reported in the UK 1946 birth cohort.9 In contrast with those data, we did not find strong evidence to suggest that the association between birthweight and diabetes differed according to adult body weight. This may be because of the small numbers of people in our cohorts who reported their birthweight, or it could also be attributable to misclassification in the recall of these measures. We have previously demonstrated that self-reported birthweight in BWHHS had a comparable distribution to that of British women of a similar age for whom there were hospital records of birthweight available and that the associations between self-reported birthweight at baseline and adult anthropometric measures were in the same direction and of a similar magnitude to what one would predict from recorded birthweights.13, 15 The recalled birthweights from the Glasgow Alumni study have not been validated. However, they are comparable with those from the 1946 birth cohort,16 although the Glasgow alumni (both men and women) reported birthweights on average 100 g heavier than the population-based sample. This could be because of the relatively affluent parents of the Glasgow alumni, who might have been better nourished than the general population.
Other studies have investigated the role that low birthweight followed by faster catch-up growth in childhood plays in the development of diabetes. Among Finnish17, 18 and Indian10 people, those who subsequently develop diabetes are more likely to have had low birthweight and to have higher than average BMI from early childhood onwards. Our results are consistent with previous observations that weight gain is an important risk factor for adult-onset diabetes.19, 20, 21 What our analysis adds is some evidence to suggest that the timing of this weight gain does not appear to be critical and that being overweight at any time in adulthood (even if there is later weight loss) is associated with increased diabetes risk. Weight gain from young to mid-adulthood was as detrimental in terms of diabetes risk as was weight gain from mid- to later adulthood. This observation underpins the importance of minimising weight gain at all ages. Our results differ from those of the US Nurses Health Study, which found that body mass index at age 18 (based on retrospective report in adulthood) was not an important determinant of diabetes risk (based on self report) once later adulthood body mass index was taken into account.8 That study is based on a single occupation group with relatively homogenous socioeconomic position (similar to the Glasgow cohort) and is of women only (similar to BWHHS). The difference between our results and those of that study may be related to the use of self-report of weight and height at all time points in the Nurses Health Study. An analysis of the British Regional Heart Study suggested that the impact of weight gain on diabetes risk was greater in those people who were thinner at baseline (when aged 40–59 years), compared to those who were heavier.19 Our results are not consistent with that finding; we found that weight gain was important in those who were and were not overweight at baseline – more consistent with the results from the Health Professional's Follow-up Study.21 However, all sets of results are consistent in the conclusion that the duration of being overweight is important, with risks of diabetes of approximately four-fold in adults who were consistently overweight compared to those who maintained normal weight throughout their lifetime. Although the importance of weight loss on diabetes risk should not be ignored,20 the primary prevention of overweight and obesity amongst children and young adults is likely to have a greater effect on the population's health, than the targeting of interventions to older adults.
In conclusion, we have demonstrated that overweight at any point in a person's life is associated with an increased risk of developing diabetes. From the time points we studied, there does not appear to be a critical time at which overweight is particularly detrimental, but the risk associated with being overweight is cumulative across the lifecourse. These results contribute to the evidence for the need to target weight control interventions to young adults.
Martyn C, Gale C, Jespersen S, Sherriff S . Impaired fetal growth and atherosclerosis of carotid and peripheral arteries. Lancet 1998; 352: 173–178.
Lithell HO, McKeigue PM, Berglund L, Mohsen R, Lithell UB, Leon DA . Relation of size at birth to non-insulin dependent diabetes and insulin concentrations in men aged 50–60 years. British Medical Journal 1996; 312 (7028): 406–410.
Eriksson JG, Forsen T, Tuomilehto J, Jaddoe VW, Osmond C, Barker DJ . Effects of size at birth and childhood growth on the insulin resistance syndrome in elderly individuals. Diabetologia 2002; 45 (3): 342–348.
Folsom AR, Kushi LH, Anderson KE, Mink PJ, Olson JE, Hong CP et al. Associations of general and abdominal obesity with multiple health outcomes in older women: the Iowa Women's Health Study. Arch Intern Med 2000; 160 (14): 2117–2128.
Gurwitz JH, Field TS, Glynn RJ, Manson JE, Avorn J, Taylor JO et al. Risk factors for non-insulin-dependent diabetes mellitus requiring treatment in the elderly. J Am Geriatr Soc 1994; 42 (12): 1235–1240.
Young-Hyman D, Schlundt DG, Herman L, De Luca F, Counts D . Evaluation of the insulin resistance syndrome in 5- to 10-year-old overweight/obese African-American children. Diabetes Care 2001; 24 (8): 1359–1364.
Fagot-Campagna A, Pettitt DJ, Engelgau MM, Burrows NR, Geiss LS, Valdez R et al. Type 2 diabetes among North American children and adolescents: an epidemiologic review and a public health perspective. J Pediatr 2000; 136 (5): 664–672.
Colditz GA, Willett WC, Stampfer MJ, Manson JE, Hennekens CH, Arky RA et al. Weight as a risk factor for clinical diabetes in women. Am J Epidemiol 1990; 132 (3): 501–513.
Hypponen E, Power C, Davey Smith G . Prenatal growth, BMI, and risk of type 2 diabetes by early midlife. Diabetes Care 2003; 26 (9): 2512–2517.
Bhargava SK, Sachdev HS, Fall CH, Osmond C, Lakshmy R, Barker DJ et al. Relation of serial changes in childhood body-mass index to impaired glucose tolerance in young adulthood. N Engl J Med 2004; 350 (9): 865–875.
McCarron P, Davey Smith G, Okasha M, McEwen J . Life course exposure and later disease: a follow-up study based on medical examinations carried out in Glasgow University (1948–1968). Public Health 1999; 113: 265–271.
Lawlor DA, Bedford C, Taylor M, Ebrahim S . Geographical variation in cardiovascular disease, risk factors, and their control in older women: British Women's Heart and Health Study. J Epidemiol Community Health 2003; 57 (2): 134–140.
Lawlor DA, Ebrahim S, Davey Smith G . The association between components of adult height and Type II diabetes and insulin resistance: British Women's Heart and Health Study. Diabetologia 2002; 45 (8): 1097–1106.
Office of Population Censuses and Surveys, Social Survey Division. The General Household Survey 1975. HMSO: London, 1978.
Lawlor DA, Davey Smith G, Ebrahim S . Birth weight is inversely associated with coronary heart disease in post-menopausal women: findings from the British women's heart and health study. J Epidemiol Community Health 2004; 58 (2): 120–125.
Hardy R, Wadsworth ME, Langenberg C, Kuh D . Birthweight, childhood growth, and blood pressure at 43 years in a British birth cohort. Int J Epidemiol 2004; 33 (1): 121–129.
Forsen T, Eriksson J, Tuomilehto J, Reunanen A, Osmond C, Barker D . The fetal and childhood growth of persons who develop type 2 diabetes. Ann Intern Med 2000; 133 (3): 176–182.
Eriksson JG, Forsen T, Tuomilehto J, Osmond C, Barker DJ . Early adiposity rebound in childhood and risk of Type 2 diabetes in adult life. Diabetologia 2003; 46 (2): 190–194.
Wannamethee SG, Shaper AG . Weight change and duration of overweight and obesity in the incidence of type 2 diabetes. Diabetes Care 1999; 22 (8): 1266–1272.
Resnick HE, Valsania P, Halter JB, Lin X . Relation of weight gain and weight loss on subsequent diabetes risk in overweight adults. J Epidemiol Community Health 2000; 54 (8): 596–602.
Koh-Banerjee P, Wang Y, Hu FB, Spiegelman D, Willett WC, Rimm EB . Changes in body weight and body fat distribution as risk factors for clinical diabetes in US men. Am J Epidemiol 2004; 159 (12): 1150–1159.
We would like to thank participants in both studies. We are grateful for funding from the Stroke Association, Chest Heart and Stroke (Scotland), NHS R&D CVD Programme and World Cancer Research Fund who funded the follow up of the Glasgow Alumni Study and the Department of Health and British Heart Foundation who funded the BWHHS study. The Centre for Public Health Research, Massey University, is supported by a Programme Grant from the Health Research Council of New Zealand. DAL and PMcC are funded by UK Department of Health public health career scientist awards. BG is partly funded through the Robert Wood Johnson Foundation. The views expressed in this paper are those of the authors and not necessarily any funding body. Part of this work was presented at the 3rd Conference on Epidemiological Longitudinal Studies in Europe (Bristol, 2004).
MJ and DAL conceived the idea, performed the analysis and drafted the first version of the manuscript. PMcC and GDS were responsible for setting up the Glasgow Alumni Study. DAL and SE were responsible for setting up the BWHHS. BG and SK are responsible for the continuing management of the Glasgow Alumni Study. All authors provided intellectual input and contributed to the final version of the manuscript.
Conflict of interest
Ethics committee approval from MREC and all relevant LRECs was received and informed consent obtained for both studies.
About this article
Cite this article
Jeffreys, M., Lawlor, D., Galobardes, B. et al. Lifecourse weight patterns and adult-onset diabetes: the Glasgow Alumni and British Women's Heart and Health studies. Int J Obes 30, 507–512 (2006). https://doi.org/10.1038/sj.ijo.0803161
- weight patterns
- weight control
- adult-onset diabetes
Association between duration of excessive weight and arterial stiffness in middle-aged and older adults
Clinical and Experimental Hypertension (2019)
Impact of body mass index and waist circumference on blood pressure: A cross-sectional survey in a population living in the Vietnam northern mountainous
Diabetes & Metabolic Syndrome: Clinical Research & Reviews (2019)
Birth Weight and Risk of Type 2 Diabetes Mellitus, Cardiovascular Disease, and Hypertension in Adults: A Meta‐Analysis of 7 646 267 Participants From 135 Studies
Journal of the American Heart Association (2018)
Pre-pregnancy weight change and incidence of gestational diabetes mellitus: A finding from a prospective cohort study
Diabetes Research and Clinical Practice (2017)
Australian Feminist Studies (2016)