Identifying Persons at Risk for Developing Type 2 Diabetes

Discussion

When comparing the classifications of diabetes risk based on questionnaire overall risk score to HbA_1c values, significant and expected increases in HbA_1c were observed as participants progressed from a risk classification of "Small" toward "Very High" or "Extreme". After collapsing the "Very High" and "Extreme" groups, the only groups that did not significantly differ were the "Moderate" and "High" risk groups. Of particular interest, those in the "Small" risk category based on the questionnaire responses had average HbA_1c values corresponding to the healthy glycemic control while those in the "Moderate" risk group had average HbA_1c values that were approaching a state of prediabetes based on the CDA diagnostic criteria. Furthermore, these "Moderate" risk individuals would be in the prediabetes range based on ADA standards which define prediabetes using an HbA_1c of 5.7–6.4%. Those in the "High" risk group, based on their questionnaire responses, had corresponding blood test scores with an average HbA_1c value at the cusp of the prediabetes classification according to the CDA range (mean HbA_1c of "High" risk group = 5.99%, CDA Range = 6.0–6.4%) and in the middle of the ADA prediabetes range (HbA1c of 5.7–6.4%) . Finally, those in the "Very High" risk group had average HbA_1c values (Mean HbA_1c = 6.6%) in the diabetes range (≥6.5%) based on both the CDA and ADA guidelines. Another related finding, with substantial clinical significance, was the extent to which the screening process identified individuals who were previously unaware of their poor glycemic control. With 75% of persons in the prediabetes range and ~62% of persons in the diabetes range based on their HbA_1c report having never been told by a physician or nurse that they had high blood sugar, serious implications regarding the need for diabetes and prediabetes screening are magnified.

Further investigation into the relationship between questionnaire outcomes and blood values using multivariate linear regression revealed that, in descending order of standardized beta values, previous diagnosis of high blood sugar (standardized β = 0.28), number of high risk parents (standardized β = 0.0.15), physical activity participation (standardized β = 0.12), age category (standardized β = 0.12), and BMI (standardized β = 0.11) were all independent significant contributors to the variability in HbA_1c. While the R statistic suggests that the model only explains 23.5% of the variance in HbA_1c, a receiver operator characteristic (ROC) analysis was performed and the area under the curve (AUC) was 0.716 using dysglycemia (HbA_1c ≥6.0%) as the primary outcome with the intention of drawing comparisons to existing diabetes risk questionnaires. The observed AUC for the ROC analysis is consistent with findings from the CANRISK (AUROC = 0.75) and FINDRISC (AUROC = 0.648 for men, 0.659 for women) questionnaires for the prediction of dysglycemia (prediabetes + type 2 diabetes). The relatively low R of this model identifies a legitimate area of further investigation to decipher what may be contributing to the remainder of the variance in HbA_1c within high risk populations. Interestingly, an analysis of the CANRISK questionnaire outcomes found that the response to their physical activity participation question was not a significant contributor to their model. This disparity between the CANRISK questionnaire and the PRE-PAID questionnaire, with respect to the significance of physical activity in the model is likely due to the fact that the PRE-PAID questionnaire had an altered version of the question which was more descriptive in its assessment of physical activity and ascertained information about physical activity frequency. These findings and the corresponding standardized beta values will be used in the future to establish weighted responses on the questionnaire with the goal of enhancing its predictive value.

The utilization of HbA_1c as the primary blood biomarker for confirmation of risk provided the investigators with a great deal of freedom in scheduling recruitment and screening events. Through the use of minimally-invasive point-of-care capillary blood testing, a broad pool of potential participants was reached. The ability to test blood in a non-fasted state and provide rapid results made this test more accessible and appealing to potential participants, thus enhancing the efficacy of recruitment efforts. The comparison between the Bio-Rad device and HPLC revealed no significant bias between the two measures which led to the decision to use the Bio-Rad samples (n = 589 with Bio-Rad values versus 304 with HPLC) for the data analysis comparing blood results to questionnaire outcomes via ANOVA and linear regression. Further, the accordance between the two HbA_1c supports the use of minimally-invasive point-of-care capillary blood testing for future type 2 diabetes and prediabetes detection initiatives that are focused on screening, awareness and education. These tests may be accessible to a larger population because they can be performed at lower costs and less intrusive to persons at risk while providing relatively accurate information, especially when used in conjunction with a risk questionnaire.

One of the primary limitations of this investigation is the demographics of the sample. In an ideal setting, a more diverse sample would provide an opportunity to enhance the validation of the PRE-PAID questionnaire for use on a broader population. In spite of this, the mandate of the PRE-PAID investigators and the funding agencies was to reach those at highest risk for developing type 2 diabetes, thus leading to more targeted recruitment efforts. The concentrated efforts aimed at reaching high risk ethnicities supports the notion that the PRE-PAID questionnaire provides a unique and appropriate tool for use in public health screening initiatives that target these populations. Another limitation of this investigation is the fact that all responses to the risk questionnaire were self-reported and several studies have shown that individuals tend to under-report their weight and waist circumference while over-reporting their physical activity habits. Although this may be a limitation, it is important to realize that during many public health initiatives, questionnaires are distributed in a similar manner and self-reported data is easier and less expensive to obtain when compared to actual measurement of the various risk factors assessed on the PRE-PAID questionnaire which may require equipment and trained personnel. Another limitation of the investigation pertains to the wording of questions assessing previous diagnoses of high blood pressure, blood sugar and family history of diabetes. Those who "didn't know" were given a score of zero. Moving forward, a more conservative approach should be taken so that those who do not know how to respond, are assumed to possess that risk factor and therefore receive a score for that question, thus contributing to their overall risk score. Finally, there have been some studies that have documented the presence of hemoglobinopathies or other conditions such as iron deficiency which would make the use of HbA_1c inappropriate for the assessment of diabetes status. The prevalence of hemoglobinopathies varies greatly depending on country and race but has been reported as high as 10% in some African populations. During the HPLC assessment of HbA_1c, no participants were identified as having hemoglobinopathies that would warrant their removal from the comparative analysis. It is possible, however, that some of the study participants who only provided Bio-Rad HbA_1c samples possessed some form of hemoglobinopathy. Additionally, there may be other factors such as prescription medication which may contribute to altered HbA_1c values and it should be noted that this data was not captured by the risk questionnaire during this study. Adding questions regarding medication use would increase the complexity and duration of completing the questionnaire which would increase subject burden.

While the strength of the CANRISK questionnaire lies in its validation using a large, and representative Canadian sample population, the PRE-PAID risk questionnaire has shown to be an effective alternative tool for use among high risk ethnicities in Canada. As a result of the PRE-PAID investigation, the CANRISK questionnaire may enhance its own predictive value if more detailed questions were included with respect to physical activity participation such as; active transport, sedentary time, physical nature of their occupation, structured exercise, leisure time physical activity plus intensity and frequency of daily activities of living. The analysis of the number of high risk parents is also unique to the PRE-PAID questionnaire which provides important information to enhance the identification of risk based on ethnicity. The ultimate goal of this investigation was to develop an inexpensive front-line questionnaire that could accurately assess a person's risk for developing diabetes.