Association is not prediction: A landscape of confused reporting in diabetes – A systematic review

• Published in: Diabetes research and clinical practice Vol. 170; p. 108497
• Main Authors: Varga, Tibor V., Niss, Kristoffer, Estampador, Angela C., Collin, Catherine B., Moseley, Pope L.
• Format: Journal Article
• Language: English
• Published: Ireland Elsevier B.V 01.12.2020
• Subjects: Association; Biomarkers; Biomarkers - blood; Clinical Medicine; Diabetes Mellitus - epidemiology; Endocrinology and Diabetes; Endokrinologi och diabetes; Humans; Klinisk medicin; Medical and Health Sciences; Medicin och hälsovetenskap; Personalized medicine; Precision medicine; Precision Medicine - methods; Prediction; Sensitivity and Specificity; Translational research; Biomarkers; Translational research; Association; Precision medicine; Personalized medicine; Prediction
• ISSN: 0168-8227; 1872-8227
• DOI: 10.1016/j.diabres.2020.108497

•Thousands of published articles report on predictive biomarkers for diabetes.•Many of these papers promise improvements in patient care.•Around 60% of these articles confuse prediction with association.•Statistically significant associations do not mean that a biomarker is predictive.•Adherence to the TRIPOD statement is encouraged for papers reporting predictive models. Appropriate analysis of big data is fundamental to precision medicine. While statistical analyses often uncover numerous associations, associations themselves do not convey predictive value. Confusion between association and prediction harms clinicians, scientists, and ultimately, the patients. We analyzed published papers in the field of diabetes that refer to “prediction” in their titles. We assessed whether these articles report metrics relevant to prediction. A systematic search was undertaken using NCBI PubMed. Articles with the terms “diabetes” and “prediction” were selected. All abstracts of original research articles, within the field of diabetes epidemiology, were searched for metrics pertaining to predictive statistics. Simulated data was generated to visually convey the differences between association and prediction. The search-term yielded 2,182 results. After discarding non-relevant articles, 1,910 abstracts were evaluated. Of these, 39% (n = 745) reported metrics of predictive statistics, while 61% (n = 1,165) did not. The top reported metrics of prediction were ROC AUC, sensitivity and specificity. Using the simulated data, we demonstrated that biomarkers with large effect sizes and low P values can still offer poor discriminative utility. We demonstrate a landscape of confused reporting within the field of diabetes epidemiology where the term “prediction” is often incorrectly used to refer to association statistics. We propose guidelines for future reporting, and two major routes forward in terms of main analytic procedures and research goals: the explanatory route, which contributes to precision medicine, and the prediction route which contributes to personalized medicine.