A large number of factors influence the relationship between obesity and diabetes and many of them are closely inter-related. That obesity on its own is not sufficient to cause diabetes is apparent from the observation that 20 per cent of patients with type 2 diabetes are not obese and even in the highest risk group with high BMI and high waist–hip ratio over 80 per cent will escape type 2 diabetes (Colditz et al.,  1995). 

Other factors include body fat distribution,  duration of obesity,  weight gain,  age,  physical activity,  diet,  the in utero environment, childhood stunting and genetic factors. Methodological issues are also important in examining the relationship between obesity and diabetes. Some of the observed increase in diabetes prevalence attributed to obesity could be related to more awareness and detection of type 2 diabetes,  rather than a true increase in numbers (previous Diabetes UK estimates are that 50 per cent of patients do not know they have type 2 diabetes).

The change in the diagnostic criteria for diabetes introduced by the American Diabetes Association in 1997, with less use of the oral glucose tolerance test and more emphasis on the fasting glucose,  appear to underestimate the prevalence of diabetes in obese subjects who may have a relatively normal fasting blood glucose in the presence of a high post-load glucose (Melchionda et al., 2002; Richard et al., 2002).

  Abdominal obesity may be an even better predictor of the development of type 2 diabetes than BMI (Larsson et al., 1984; Ohlson et al., 1985). The predictive value of high waist–hip ratios and high waist circumferences in mediating the risk of diabetes appears to be of most importance in those in the highest quintile of these measures (Chan et al.,  1994)  and perhaps in leaner subjects.

For a given BMI many Asian populations have a much higher risk of type 2 diabetes, even at a BMI well within the normal range (Mather and Keen, 1985).

WHO is currently studying the use of a more limited range of normal BMI (18.5–22.9 kg m^-2)  in these groups together with use of waist circumference (James et al., 2001). The duration of exposure to obesity is an important modulator of the risk of diabetes. In Pima Indians, subjects whose BMI was greater than 30 kg m^-2  for more than 10 years had over twice the risk of type 2 diabetes compared with those who had been obese for less than 5 years (Everhart et al.,  1992).  The epidemic of childhood obesity allied with the influence of obesity duration suggests both increasing frequency of diabetes and its earlier onset.

Weight gain during adult life acts in addition to BMI per se to modify the risk of diabetes. In the Health Professionals’ Follow-up Study men who gained more than 13.5 kg over the five years of the study had a 4.5-fold increased risk of diabetes in comparison with those men who remained within 4.5 kg of their weight at entry to the study (Chan et al., 1994). Similar findings apply to women as described in the Nurses’ Health Study where the relative risk of diabetes was 2.7 in those who gained 8–10.9 kg compared with those who were weight stable over a 14-year period (Colditz et al., 1995).

Alongside the epidemic of childhood obesity and diabetes there is also an epidemic of diabetes related to the ageing population. The prevalence of type 2 diabetes increases progressively with age peaking at 16.5 per cent in men and 12.8 per cent in women at age 75–84 years (Wilson and Kannel, 2002). Obesity rates plateau about 20 years earlier (Prescott-Clarke and Primatesta, 1999), but the age-related increases in total body fat and visceral adiposity make BMI a less good marker of adiposity in older age groups. Indeed, many normal weight elderly men and women are at high risk of type 2 diabetes due to increased visceral abdominal fat (Goodpaster et al., 2003). In the UK it is projected that due to population ageing by 2036 there will be 20 per cent more cases of type 2 diabetes than in 2000 (Bagust et al., 2002).

Decreasing levels of physical activity are undoubtedly implicated in the epidemic of obesity,  but physical activity also has independent protective effects on the risk of diabetes. In the British Regional Heart Study, whilst BMI was the dominant risk factor for diabetes, men engaged in moderate levels of physical activity had a substantially reduced risk of diabetes,  relative to the physically inactive men, even after adjustment for age and BMI (Perry et al., 1995). Similar data in women demonstrate a relative risk of type 2 diabetes of 0.67 in those who engaged in vigorous exercise at least once a week compared with women who did not exercise weekly (Manson et al., 1991). At least in terms of reducing the risk of type 2 diabetes it is probably better to be overweight and physically active, than to be normal weight and inactive (Wei et al., 1999).

Dietary factors appear to have effects independent of those on obesity on the development   of type 2   diabetes.  Increasing   fat in the diet is associated with both obesity and the development of diabetes (West,  1978),  but much of this link is explained simply by the high energy intake that accompanies high fat diets. However, some populations with high-fat diets (e.g. Eskimos and the Japanese)  have a relatively low prevalence of diabetes compared with that expected from their obesity rates and this may be explained by a high intake of omega-3 polyunsaturated fatty acids (Malasanos and Stacpoole,  1991).  A recent large prospective study of diet in women aged 34 to 59 years without diabetes at baseline and followed for 14 years found that total fat intake was not associated with risk of type 2 diabetes,  but for a 5 per cent increase in energy from polyunsaturated fat, the relative risk was 0.63 and for a 2 per cent increase in energy from trans fatty acids the relative risk was 1.39 (Salmeron et al.,  2001).  The authors estimated that replacing energy derived from trans fatty acids with polyunsaturated fat would lead to a 40 per cent lower risk of type 2 diabetes.

Whilst the vast majority of studies either show diabetes rates rising as obesity prevalence climbs, or project such a rise in diabetes from the observed obesity prevalence, one Swedish population survey has not demonstrated this (Eliasson et al., 2002). From 1986 to 1999 the mean BMI in adults in northern Sweden increased from 25.3 to 26.2 kg m^-2      and the prevalence of obesity rose from 11 to 15 per cent. However, in spite of the marked increase in obesity there was no increase in the prevalence of known diabetes. Dietary factors may account for some of this discrepant finding,  with the diet over this period containing less saturated fat and having a lower glycaemic index.  One additional observation was a decrease   in waist–hip ratio (representing   reduced   visceral   adiposity), perhaps also contributing to the absence of a BMI effect on diabetes. Findings such as these may mean that the gloomy picture of the diabetes epidemic painted by many authors is not quite as inevitable an outcome as projected.

In contrast, data from Australia show a dramatic increase in diabetes over the last 20 years, with a doubling of diabetes prevalence to its current value of 7.4 per cent (Dunstan et al., 2002). An additional 16.4 per cent had abnormal glucose tolerance. Although obesity rates in this population have increased, neither obesity nor changes in the age profile of the population fully explain the extent of the diabetes epidemic.  It is likely that some of the other factors discussed above, including body fat distribution, duration of diabetes and physical activity account for this adverse pattern.

Whilst obesity is clearly important, other factors appear to influence the susceptibility both to weight gain and to the development of diabetes. The ‘thrifty’ gene hypothesis (Neel,  1962)  suggests that the obese-type 2 diabetes mellitus genotype may have had some survival advantage, perhaps by favouring fat storage at times when food was abundant, so leading to improved survival during famines. However, this hypothesis remains an epidemiological explanation, with the exact genetic factors remaining unclear and no prospective data showing a survival advantage in subjects felt to have a thrifty genotype. Much recent discussion has centred on the potential importance of the in utero environment in the causation of later type 2 diabetes (Hales and Barker, 2001).

This hypothesis, the ‘thrifty’ phenotype hypothesis suggests that the epidemiological associations between poor fetal and infant growth and the subsequent development of type 2 diabetes results from the effects of poor nutrition in early life, which produces permanent changes in glucose and insulin metabolism.  These changes lead to reduced insulin secretion and increased insulin resistance and hence predispose to type 2 diabetes. The relative contribution of thrifty genes, a thrifty phenotype and later environmental factors (e.g. physical activity and diet) to obesity-related diabetes is not clear. 

Perhaps the trend for average birth weights to gradually increase in many populations indicates that the in utero environment is improving and this may mean reduced susceptibility to later diabetes.  However,  the explosion in childhood obesity is likely to outweigh the potentially beneficial birthweight effect in most populations.

One further adverse early life effect on susceptibility to obesity and probably also to later diabetes is childhood stunting. This remains common in lower income countries. Surveys in a number of countries show a significant association between stunting and overweight status in children of all countries (Popkin et al.,  1996).  With the transition from a lower income developing country to a more affluent developed one it is likely that the increased early stunting- mediated susceptibility to obesity will be compounded by later economic and social changes driving up obesity and diabetes rates.

Conclusions
There is no doubt that obesity is at present the major player in the increasing prevalence of type 2 diabetes. The current epidemic of obesity shows little sign of abating in most parts of the world and in contrast is still accelerating, particularly in children.  Global predictions of the diabetes epidemic with 300 million patients with type 2 diabetes by 2025 are well on course (King et al., 1998). Many of the factors that modify the relationship between obesity and diabetes, such as duration of obesity and physical activity levels are also changing adversely and are exacerbating the diabetes epidemic. The challenge for society is to reverse the ever-increasing prevalence of obesity.

Anthony H. Barnett
  Department of Medicine, University of Birmingham and Birmingham Heartlands and Solihull NHS Trust (Teaching), UK

Sudhesh Kumar
Professor of Medicine, Diabetes & Metabolism, Warwick Medical School, University of Warwick, UK