Evaluating Diagnostic Models for Undiagnosed Type 2 Diabetes

Overview of the Systematic Review

A systematic review has been initiated to evaluate two significant clinical prediction models - the Leicester Risk Assessment score (LRA) and the Leicester Practice Risk score (LPR). These models are crucial for detecting undiagnosed type 2 diabetes and prediabetes, affecting millions across the UK.

Public Health Context

In the UK, it is estimated that around one million adults live with undiagnosed type 2 diabetes, in addition to 5.1 million suffering from non-diabetic hyperglycaemia. This condition, where blood glucose levels are elevated but not high enough for a diabetes diagnosis, can lead to severe complications, reducing life expectancy by nearly ten years.

The implications of untreated diabetes extend beyond personal health, representing a significant economic challenge, with 10% of the NHS's annual budget directed toward managing type 2 diabetes.

Understanding the Leicester Risk Scores

Developed over 15 years, the LRA employs a straightforward 'tick box' questionnaire to generate a score based on various factors like age, gender, family history of diabetes, waist size, and body mass index. This tool is recommended by the National Institute for Health and Care Excellence (NICE) for high-risk adults, aiding in timely intervention through programs like the NHS Diabetes Prevention Programme.

The LPR, a variant of the LRA, is adapted for primary care settings and notably excludes waist circumference due to inconsistent recording practices in medical records. While initially designed as diagnostic tools, these models have also evolved into prognostic instruments, assisting in predicting future diabetes risk.

Methodology of the Review

The primary objective of this systematic review is to assess both the LRA and LPR's effectiveness in identifying individuals at risk of undiagnosed type 2 diabetes. The researchers plan to conduct a meta-analysis to summarize the accuracy of these models across varied populations by analyzing data from previously published validations and updates.

Instead of traditional database searches, a citation-based strategy will be employed. By searching platforms like Scopus, Web of Science, and Google Scholar, the review aims to capture all relevant studies concerning these models since August 2010.

Data Extraction and Analysis

The review team, consisting of two independent reviewers, will extract data related to study design, patient demographics, and model performance metrics. Key performance measures will include the calibration slope, intercepts, and the c-statistic, essential for determining the models' ability to differentiate between those with and without diabetes. If five or more studies provide sufficient data, a random-effects meta-analysis will be conducted.

To enhance reliability, the study will also consider any potential biases in included studies, assessing risks associated with participant selection and modeling approaches. Additionally, funnel plots will be used to check for publication bias.

Challenges and Limitations

One limitation identified in the LRA is the absence of reported intercept values in its original publication, which complicates the validation process. The review will also discern whether validations involved recalibration efforts, as this distinction holds significant importance for interpreting results in new populations.

Conclusion and Future Implications

By focusing on variations in model performance across distinct populations and healthcare systems, the findings from this systematic review could lead to improvements in early diagnosis and management of undiagnosed diabetes. The hope is that a clearer understanding of these models' applicability and effectiveness will foster better health outcomes for individuals at risk for diabetes.