Methodology for the estimates on diabetes and impaired glucose tolerance
1. Data search
The search for data was limited to studies published after 1979. This cut-off was chosen as data collected prior to 1980 may no longer reflect the current prevalence of diabetes. Selection of articles was limited to those published pre-March 2006.
The Medline database and internet were used for the literature search. Systematic searches were conducted for each country using the following search formulae:
- Country name (all the countries of the world were entered for separate searches) together with ‘diabetes’ or ‘impaired glucose tolerance’ and ‘prevalence’ or ‘incidence’; and
- ‘NIDDM’ or ‘IDDM’ or ‘non-insulin-dependent diabetes mellitus’ or ‘insulin-dependent diabetes mellitus’ or ‘Type 1 diabetes’ or ‘Type 2 diabetes’, combined with ‘prevalence’ or ‘incidence’.
Relevant citations from each article were also obtained. A number of other avenues were explored in the search for relevant data. Diabetes researchers in each major IDF geographical region were contacted and requested to provide information on the prevalence of diabetes for countries within their region. In addition, IDF member associations in each member country were asked about relevant data. In the absence of data for a country, the member association was further asked to comment on the use of data from another country (see section on Extrapolation below).
2. Data selection
The search obtained data in a variety of forms such as prevalence studies, registry reports, hospital statistics, government estimates, etc. Studies for a particular country were included based on their level of reliability. The following factors were taken into account when assessing a study’s level of reliability:
- The year of the study — more recent studies were preferred.
- The screening method used — the oral glucose tolerance test (OGTT) was the preferred method of screening, followed by two-hour blood glucose (2hBG) alone, then the fasting blood glucose (FBG) alone, and then self-report (SR).
- Sample size — studies with larger sample sizes and higher response rates were preferred.
When more than one study was available for a country, and there was no clear superiority of one over the other, the results from the available studies were averaged, and then applied to the national population.
If there were no data available for a particular country, prevalence rates from a published study from the socio-economically, ethnically, and geographically most similar country were applied to that country’s age and sex-specific (and in the case of low/middle-income countries, urban/rural-specific) population distribution. Socio-economic comparisons were based on gross national product (GNP) per capita. Ethnic comparisons were based on ethnicity data from the CIA World Factbook 2005 1 .
If a dataset did not provide sex-specific data, the data were disaggregated and assigned 50% to females and 50% to males.
Urban: rural prevalence
In countries with low or middle-income economies, differences between urban and rural populations in levels of physical exercise, diet, and socio-economic factors often result in significant differences in diabetes prevalence rates. Therefore, for low- and middle-income economies (except those of the former socialist economies in Europe), the urban and rural rates were calculated and numbers reported separately.
The economies were defined according to the 1997 GNP per capita, calculated using the World Bank Atlas method 2 . Low- and middle-income economies had a GNP per capita of less than USD9,655, and high-income economies had a GNP per capita of USD9,655 or more. If the above conditions for different urban and rural diabetes prevalences applied, then for countries where available studies showed prevalences separately for urban and rural populations, these rates were applied to the national urban and rural populations.
For studies reporting on a mixed urban and rural population, but where no data were provided as to the urban/rural distribution of the survey population, the available age and gender specific data were assigned to the population so as to produce a 2:1 urban:rural ratio in diabetes prevalence.
For countries where only urban or only rural data were available, the 2:1 ratio was used to calculate the prevalence of diabetes in the other segment of the population. No urban:rural difference was used for IGT prevalence, unless the data for that country indicated a prevalence difference to be present.
Studies from several countries — Canada, France, Germany, Israel, Italy, Netherlands, New Zealand, Norway — only provided data on self-reported diabetes. To account for undiagnosed diabetes, the prevalence of diabetes for Canada was multiplied by a factor of 1.5, in accordance with findings from the USA 3 , and for the other countries doubled, based on data from a number of countries 4 5 6 7 8 .
3. Prevalence calculation
A list of the world’s countries and 2007 and 2025 population distribution estimates was obtained from the United Nations Population Division 9 . The age- and sex-specific prevalence rates (obtained from the logistic regression, see below) were applied to the corresponding age and sex population distribution for the years 2007 and 2025 for each country. This method for estimating figures for 2025 only takes into account changes in age, sex and urban/rural population distributions, and not for the likely changes in lifestyle and obesity, which may tend to increase diabetes prevalence. Thus, the figures may be an underestimate.
The prevalence rate (PR) of diabetes and IGT for each country was then calculated using the formula:
|PR (for those people 20-79 years) =||Total number of expected cases (20-79)
|Total country population (20-79)|
Total number of expected cases of diabetes, or IGT, in the 20-79 year range = the sum of each age and gender (and urban/rural) specific number, as derived according to the earlier description.
Following calculation of the PR, the expected number of people with diabetes and IGT within the country was reported separately for males and females, according to age groups (20-39, 40-59, 60-79), and in those low- and middle-income economies (only for diabetes), according to residence in urban and rural areas.
For countries without available age and gender distribution descriptions i.e. those with populations of less than 100,000 for the year 2000, (and Taiwan), for which data are not provided 9 , the total world population distribution was applied to the 2005 population as indicated in the CIA World Factbook 2005 1 . For Andorra, Liechtenstein, Monaco and San Marino, the total developed world population was applied. Populations for all these countries for 2007 were obtained by applying the annual increase for one year, and for 2025, by assuming an unchanged proportion of the world (or developed world) from 2005 to 2025.
The countries/territories without UN population data that are included are: Andorra, Anguilla, Antigua and Barbuda, Aruba, Bermuda, British Virgin Islands, Cayman Islands, Dominica, Grenada, Cook Islands, Kiribati, Liechtenstein, Marshall Islands, Monaco, Nauru, Niue, Palau, Saint Kitts and Nevis, San Marino, Seychelles, Taiwan, Tokelau, Tuvalu.
4. Prevalence reporting
In addition to calculating the national rates, a prevalence for each country and region, adjusted to the world population, was calculated by applying for each country that country’s age- and sex-specific rates to a notional population of that country’s population size, but with the world population age and gender distribution for 20–79 years (for 2007 and 2025). This was done to facilitate comparison of rates between countries and regions, and this adjustment to the world population noted whenever it was used.
For each region the prevalence adjusted to the world population was calculated by the summation of the number of persons for each member country with the condition, if each country’s world population adjusted prevalence were applied to that country, and the sum divided by the total regional population (20-79 years).
5. Logistic Regression
For each country, data for both diabetes and IGT are presented for people in the 20-79 age group. Most of the datasets used did not contain data for all age groups in the 20-79 year age bracket. In order to fill in missing data and to ensure a smooth relationship between prevalence and age, logistic regression was performed on those datasets that contained four or more datapoints.
Observed data were entered into an SPSS spreadsheet under the following columns: age (mid-age of each age group), weight (number of people without or with diabetes, or IGT, for each age group), and diabetes or IGT (0 = no, 1 = yes). The age specific prevalence (or case numbers, when provided) was used to obtain the weighting in the following manner:
|If 3.6% of 1,000 participants of a particular age group had the condition (diabetes or IGT), the weighting for having the condition would be 36, and for not having the condition, 964.|
Following this, the variable age2 (age x age) was calculated, to enable the model to contain a quadratic term, so that the end model could include the possibility of flattening or reducing prevalence for the oldest age groups. A binary logistic regression was then performed using diabetes or IGT as the dependent variable and age and age2 as the covariates, to produce parameter estimates for the intercept, B and C. This provided the values for each of the 12 five-year groups (20-24, 25-29, …75-79) for the following equation:
|y = Intercept + (B x age) + (C x age2)|
The age specific prevalence (for the five-year age group) was then calculated as (ey/(ey+1)).
The total numbers of persons with diabetes and IGT for each country were then calculated by applying the calculated age specific prevalence rates to the demographic data from the United Nations Population Prospects 9 .
An upper limit of age was necessary for the logistic regression process, and 79 years was the limit chosen. When original datasets contained the age group 65+, the assumption was made that this age group was 65-74. If a dataset contained the age group 60+, the assumption was that this age group was 60-79, unless all previous age group data were in 10-year groups, in which case a 60-69 year limit was applied. No age groups with the youngest members being over 79 years were included, but persons over 80 years were included if part of an age group 75-84 years.
Where the data were available, five-year age bands were chosen instead of 10-year age bands as they provided 12 datapoints in the 60 years age range which gave a smoother relationship between age and diabetes prevalence.
The following figures illustrate how the published age specific data could be converted by using the described methodology into a smoothed curve with respect to age.
Examples of modelled and published diabetes prevalences
Figure A1.1 Jordanian males and females combined (urban)
Figure A1.2 Chinese males
Figure A1.3 Bolivian females
Figure A1.4 Indian females
1.Central Intelligence Agency The World Factbook.CIA..Accessed 0 2005
2.The World Bank Development Report - Knowledge for Development 1998/99. New York, USA: Oxford University Press; 1999
3.Harris,M.I. Flegal,K.M. Cowie,C.C. Eberhardt,M.S. Goldstein,D.E. Little,R.R. Wiedmeyer,H.M. Byrd-Holt,D.D. Prevalence of diabetes, impaired fasting glucose, and impaired glucose tolerance in U.S. adults. The Third National Health and Nutrition Examination Survey, 1988-1994. Diabetes Care.1998; 21(4): 518-524
4.Gourdy,P. Ruidavets,J.B. Ferrieres,J. Ducimetiere,P. Amouyel,P. Arveiler,D. Cottel,D. Lamamy,N. Bingham,A. Hanaire-Broutin,H. Prevalence of type 2 diabetes and impaired fasting glucose in the middle-aged population of three French regions - The MONICA study 1995-97. Diabetes Metab.2001; 27(3): 347-358
5.Rathmann,W. Haastert,B. Icks,A. Lowel,H. Meisinger,C. Holle,R. Giani,G. High prevalence of undiagnosed diabetes mellitus in Southern Germany: target populations for efficient screening. The KORA survey 2000. Diabetologia.2003; 46(2): 182-189
6.Mooy,JM Grootenhuis,PA de Vries,H. Valkenburg,HA Bouter,LM Kostense,PJ Heine,RJ Prevalence and determinants of glucose intolerance in a Dutch Caucasian population. The Hoorn study.. Diabetes Care.1995; 18: 1270-1273
7.Eliasson,M. Lindahl,B. Lundberg,V. Stegmayr,B. Diabetes and obesity in Northern Sweden: occurrence and risk factors for stroke and myocardial infarction. Scand J Public Health Suppl.2003; 61: 70-77
8.Dunstan,D.W. Zimmet,P.Z. Welborn,T.A. de Courten,M.P. Cameron,A.J. Sicree,R.A. Dwyer,T. Colagiuri,S. Jolley,D. Knuiman,M. Atkins,R. Shaw,J.E. The rising prevalence of diabetes and impaired glucose tolerance: the Australian Diabetes, Obesity and Lifestyle Study. Diabetes Care.2002; 25(5): 829-834
9.United Nations Population Division World Population Prospects: The 2004 Revision. Geneva: United Nations; 2005