Epidemiology of diabetes mellitus
The epidemiology of diabetes is a discipline that came into its own in the second half of the 20th century. Its role is to guide the definition and classification of diabetes, to provide quantitative estimates of the distribution, manifestations and impact of the condition at a population level (descriptive epidemiology), to assess and analyse the role of risk factors in development of the disease and its consequences (analytical epidemiology), and to design and evaluate clinical trials or develop and test hypotheses (experimental epidemiology). Diabetes was once considered a rare disease, but rose to global epidemic proportions by the end of the twentieth century. This change has undoubtedly been driven by increasing affluence and widespread adoption of "western" lifestyles and diet, but many other factors need to be taken into account. The task of diabetes epidemiology is to describe and analyse the causes and consequences of the pandemic of type 1 and type 2 diabetes, and to develop and test strategies that might help to put this pandemic into reverse. This page provides a brief overview of a modern epidemic and the discipline that has arisen to combat it.
The epidemiology of diabetes is a relatively recent development. Kelly West, the founding father of the discipline, points out that the word "epidemiology" is not mentioned in the 1959 edition of the Joslin textbook, and relates that he himself "did not even realize his work constituted epidemiology until he was informed of this by a 'real' epidemiologist in 1966!"
Scope of Epidemiology
West considered that there were 4 main elements to the epidemiology of diabetes:
- Definition and classification of diabetes.
- Descriptive epidemiology, embracing the "distribution of diabetes or its manifestations in relation to time, place, person, and the natural history of the disease".
- Analytic epidemiology, including the study of risk factors, their relation to the disease and their aetiological significance.
- Experimental epidemiology, including design and evaluation of trials of prevention or treatment, and the development of experimental designs to teats specific hypotheses.
The beginnings of diabetes epidemiology
Diabetes was once considered a rare disease. William Osler's 1892 textbook of medicine devoted 10 pages to diabetes, as against 65 to tuberculosis. If diabetes was rare in the nineteenth century, childhood diabetes was truly exceptional. The Massachusetts General Hospital admitted 47,899 patients from 1824-98, of whom 172 (0.004%) had diabetes. Of these, 18 were diagnosed under the age of 20, and 3 under the age of 10. When John Lovett Morse, Professor of Pediatrics at the Harvard Medical School, wrote the first paper on childhood diabetes in English in 1913 he had personal knowledge of only 19 cases.
Elliot Joslin (1870-1962) was among the first to apply statistical analysis to the study of diabetes, initially with the help of data collected for purposes of life insurance. West credits Joslin with the first systematic recognition of the importance of obesity in the development of diabetes.
Risk Factor Disease
One of the central tasks of epidemiology is to describe the pattern of disease within a population, to identify those at increased risk (for example, workers in certain occupations), and to analyze possible reasons for differences in risk (cigarettes and cancer). Identification of risk factors associated with disease opens the way to experimental epidemiology, in which potential risk factors are modified prospectively.
To take one example, the suggestion that cow's milk might be associated with subsequent development of type 1 diabetes has been tested in a multinational trial in which neonates are randomly assigned to formula feeds in which cow's milk is present or absent following weaning. See Cow's milk and other early nutritional influences
Prospective studies of populations, such as the Framingham study, came into their own after World War 2, and allowed risk factors for the development of cardiovascular disease and diabetes to be identified, more especially the triad of hypertension, hyperlipidaemia and hyperglycaemia, all of which are associated with central obesity and hyperinsulinaemia. This constellation is known as the metabolic syndrome.
With the rise of risk factor disease, the emphasis of diabetes management moved away from palliation towards prevention. Prevention relies on population-based assessment of risk, and the epidemiologist has in consequence played an increasingly prominent role in collecting and evaluating evidence and applying this to management guidelines.
Glucose as a risk factor
Early clinical management of diabetes was based upon the coexistence of symptoms and glycosuria. Asymptomatic hyperglycaemia was recognised but generally not considered worth treating. Meanwhile, there was intense controversy as to the clinical utility of glucose control designed to achieve near-normoglycaemia as against mere symptom control. Epidemiological studies played a key role in formulating our current definition of diabetes, based around detection of a glucose threshold for diabetic retinopathy.
These studies demonstrated a clear threshold effect for retinopathy, but a continuously distributed risk for arterial disease; the association between HbA1c and heart disease, for example, extends well into the non-diabetic range (see Complications of diabetes . An operational decision had nonetheless to be made as to the glucose level at which intervention to lower the risk of heart disease would be justified, and this empirical cut-off between health and diabetes was designatedImpaired Glucose Tolerance (IGT). Parallel definitions of intermediate hyperglycaemia in terms of fasting glucose and HbA1c were subsequently introduced, but the groups identified show surprising little overlap with IGT, and this remains a controversial area.
The observational studies linking hyperglycaemia to small and large vessel disease led to large-scale randomized trials of interventions designed to reduce the risk of vascular complications in diabetes. Of these, the Diabetes Control and Complications Trial (DCCT) established beyond doubt that near-normal glucose control can delay or prevent microvascular complications. A series of major trials in type 2 diabetes have however failed to show that improved glucose control improves survival or reduces the morbidity from heart disease.
This implies that there is an important qualitative difference between the glucose thresholds for diabetes and IGT. Small vessel complications are unifactorially linked to glucose levels, respond to glucose-lowering therapies, are sharply inflected above the diagnostic threshold, and are reproducible in widely differing populations. In contrast, the risk of macrovascular disease is multifactorial (Hypertension and hyperlipidaemia help to modulate the risk) , responds poorly to glucose lowering therapies, shows a smooth increase from within the non-diabetic population, and varies markedly from one age group or ethnic group to another. No surprise that the diagnostic threshold for diabetes has never been challenged, whereas the threshold for macrovascular intervention has given rise to endless controversy.
The pandemic of diabetes
Descriptive epidemiology has traced the dramatic rise of both forms of diabetes in recent decades:
Type 1 diabetes: Type 1 diabetes (or at least diabetes in children) was rare before the 1950s, and its incidence appears to have risen steadily in populations of European extraction since then, with a doubling time across Europe as a whole of 20-25 years. More recently, a similar trend has been observed in many other populations around the world. Type 1 diabetes is an immune-mediated condition whose causes are still unknown, and which is discussed in more detail in the section on Epidemiology of type 1 diabetes.
Type 2 diabetes: Type 2 diabetes has typically appeared first in the more affluent members of the population, an observation that dates back to the ancient Sanskrit writings. The twentieth century has witnessed two major developments. The first is that, within western societies, diabetes spread from the more wealthy to less advantaged social groups, and then became more prevalent among the poor than among the better-off. The second is that increasing affluence and adoption of a western lifestyle has produced a rapid increase within traditional societies in which it was previously rare. This is sometimes referred to as transitional diabetes, and once again has first affected the more affluent. Thus it might be said that diabetes is a disease of the poor in rich countries, and of the rich in poor countries. See Epidemiology of type 2 diabetes.
Methods of Investigation
Observational studies set out to examine the undisturbed natural history of disease or its treatment. This can be done retrospectively or prospectively, but in the latter case participants should ideally be unaware that they are taking part in a study, since this might in itself influence their behaviour. The rise of computerized medical records has made it increasing possible to analyze real-life health data, but this approach is not without its pitfalls, discussed elsewhere in Diapedia. Observational studies may be structured in different ways:
- Case series, as the name implies, collect and review the characteristics of similar patients in a relatively unstructured way. Simple comparisons may then emerge, exposure to the same drug or industrial environment, for example, but the case series is generally no more than the first step in a lengthy process.
- Case-control studies are more powerful, because they allow matched comparisons between people with and without a given disease. For example, people with lung cancer can be compared with people without cancer who share the same environment and characteristics. Exposure to cigarettes can then (for example) be compared in the two groups.
- Cohort studies focus upon exposure rather than disease. A cohort study might for example be based upon retrospective or prospective study of people selected for exposure or non-exposure to asbestos, and would then go on to identify differences in their health outcomes which might have been influenced by asbestos.
Once a risk factor for disease has been identified, it may be so powerful that further analysis is unjustified. No-one, for example, would wish to undertake a controlled trial of cigarette smoking or asbestos exposure. Risk factors rarely emerge so clearly, however, and confounders may often be present. A confounder is an unmeasured variable that might sway the result. To take one example, observational studies showed that women on hormone replacement therapy (HRT) were healthier and lived longer. The observation seemed so clear that a randomized trial was considered unnecessary or even unethical. When such a study was however performed, it soon emerged that HRT was harmful. The impression of benefit had arisen because women who were previously offered HRT tended to be more affluent and health conscious than those who were not - a major confounder.
Randomized controlled trials (RCTs) therefore offer the highest level of evidence, because they are designed as an experiment in which the patients are selected at random, and (ideally) neither they nor their doctors know which treatment they are receiving. RCTs are not infallible, however, and commercial trials of new therapy can easily be set up in such a way as to bias the outcome. See Interpreting Clinical Trials.
^ West KM. Epidemiology of diabetes and its vascular complications. Elsevier, New York, 1978.