Teng A. Diabetes Res Clin Pract. 2019 Jan 8. pii: S0168-8227(18)31175-6.
Rates of progression from pre-diabetes to diabetes internationally range from around 2-18% per year. HbA1c has become an acceptable and common first-line test for diagnosis of pre-diabetes and diabetes internationally yet progression of prediabetes defined by HbA1c has rarely been studied. Accurate pre-diabetes progression rates are important for planning and monitoring interventions to prevent diabetes in people with pre-diabetes. Differences in diabetes incidence by ethnicity are common internationally; however, more information is required to identify the key policy levers and interventions to address ethnic disparities. Teng A, et al., conducted a study to examine the incidence of diabetes in an ethnically diverse primary care population with newly diagnosed pre-diabetes; and second, to quantify the factors that protect against progression to diabetes, with an emphasis on non-clinical as well as clinical factors.
Method
An enrolled population of 130,000 was selected as the source population by a primary health organization. Eligible individuals were aged 25 years and older, with a first-time reported diagnosis of pre-diabetes (HbA1c 5.9-6.6%) and at least one subsequent HbA1c test result. Individuals were followed up from the time they were first diagnosed with pre-diabetes (baseline) to determine whether they got type 2 diabetes (HbA1c 6.7%+/≥ 50 mmol/mol) in a cohort study design. HbA1c was used for diabetes diagnosis because it was the recommended first-line test for diagnosis of type 2 diabetes during the study period. Ethical approval was awarded by the Otago Human th Ethics Committee (HD16/002, 20 September 2016). In the Primary health organisation patients sign to agree to the use of their health information for research as approved by an ethics committee. There were three sources of variables that were linked: variables recorded in primary care records, health-related variables recorded in national datasets held by the Ministry of Health, and socio-demographic variables held in other national datasets in the Integrated Data Infrastructure (IDI). The IDI is a government-led research database of administrative, survey and non-government data. HbA1c test results and Body Mass Index (BMI) measures were generated from primary care records. BMI records were selected if recorded in the 5 years before baseline or 30 days after, and were categorized into standardized groupings. Ministry of health datasets were used to identify antidepressant use (five years prior to baseline), most recent primary care consultation at baseline, existing diabetes and other chronic conditions (within the five years prior to baseline), metformin treatment during follow-up (pharmaceutical dataset), bariatric surgery, and hospital admission with diabetes/hyperglycemia in pregnancy. Kaplan Meier survival analysis estimates diabetes prevalence at 3 years post-diagnosis were produced for each ten year age group and by sex, ethnicity, baseline HbA1c, education, deprivation, household income and time period. These analyses were done on the full available dataset. Proportional hazards (Cox) regression models were used to estimate the association between variables of interest and pre-diabetes progression. Analyses were run on the ‘non-missing dataset’ that had no missing data on education, personal income or deprivation. These socio economic position (SEP) variables were the only variables in the full model that had missing data. A sensitivity test was done to examine the effect of excluding records with missing SEP data. A model with adjustment for age, sex, ethnicity and HbA1c was run for each variable of interest on the full set of available data.
Results observer were
- From a total of 65,802 individuals, 14,043 individuals with pre-diabetes met the inclusion criteria, with at least one HbA1c record
- The study group was followed up for a total of 432,645 person months to calculate diabetes incidence
- Regression analyses were carried out on the two-thirds of individuals (n=9,222) in the ‘non-missing dataset’
- The cumulative incidence of diabetes was 1.04% at one year, 2.99% at two years, 4.95% at three years, 7.65% at four years and 9.45% at five years. Progression was 5.8% for men and in women it was 4.2%
- Women experienced one-third lower progression rates than men
- By age, 35-44 year olds showed the highest progression rates, who had nearly three times higher diabetes incidence than 65+ year olds, who had the lowest progression rates
- Rates of progression were similar by ethnicity after adjusting for HbA1c status
- There was a strong protective association between speaking te reo (Indigenous language) and progression to diabetes
- Glycated hemoglobin at baseline was strongly (p<0.001) and independently associated with progression to diabetes with a clear dose-response relationship
- BMI was independently associated (p=0.033) with progression to diabetes with a dose-response relationship
- There was weak (p=0.067) evidence that smoking and ex-smoking were associated with greater diabetes incidence compared to never smoking
- Metformin treatment was associated with an 88% decreased progression to diabetes
Thus, it was concluded that the contemporary study of a pre-diabetes cohort demonstrated lower rates of diabetes progression than expected. Indigenous language, lower BMI and never smoking were protective against pre-diabetes progression and are targets for prevention policies.
