Systematic review of 88 observational studies on missing data reporting in T2DM-MCI research

Photo by Brett Jordan / Unsplash

Frontiers in Medicine Published May 11, 2026 Medically reviewed May 11, 2026 DOI ↗ By Dr. Amelia Tan, PhD · Internal Medicine & Chronic Disease

Key Takeaway

Note that missing data handling in T2DM-MCI research is often insufficient and may introduce bias.

This systematic review evaluated reporting and handling of missing data across 88 observational studies related to Type 2 Diabetes Mellitus with Mild Cognitive Impairment. The review included 78 studies in English and 10 in Chinese. The authors assessed the extent of missing data and the methods used to address it.

The analysis found that only 22.7% (n = 20) of the studies quantified the missing data, with an average of 9.1%. Among the studies with missing data (n = 23), 52.2% (n = 12) provided reasons, primarily poor quality of data collection (41.7%) and loss to follow-up (41.7%).

Complete case analysis was the predominant method for handling missing data, used in 93.3% of cases. Only 4.4% (n = 1) performed a sensitivity analysis. The authors highlight that reporting of missing data remains ambiguous and methods employed to handle missing data are insufficient, which may potentially introduce bias.

The review assesses reporting and handling of missing data, not clinical outcomes. Research outcomes post-COVID-19 pandemic indicate a rebound, with China maintaining a leading position in scientific research output. Clinicians should interpret these findings as descriptive of research methodology rather than clinical efficacy.

Study Details

Study typeMeta analysis

EvidenceLevel 1

PublishedMay 2026

View Original Abstract ↓

BackgroundMissing data is common in observational studies, and even more so in type 2 diabetes mellitus with mild cognitive impairment(T2DM-MCI), which limits the completion of assessments. We evaluated the extent, current reporting, and handling of missing data, as well as the prevailing research trends in observational studies related to T2DM-MCI.MethodsA systematic search of PubMed, Embase, and Cochrane Library was conducted from January 2020 to April 2025 to identify observational studies related to T2DM-MCI. Bibliometrics was performed using VOSviewer and CiteSpace to evaluate publishing trends, authors, journals, and keywords. The reporting and handling of missing data were assessed according to the guidelines recommended by STROBE and Sterne et al., with a focus on the recording, causes, mechanisms, processing methods, and sensitivity analysis of missing data. Data analysis was conducted using SPSS 26, and visualization was performed using Origin Pro 2024.ResultsAmong the 4,471 screened records, 88 studies (78 in English and 10 in Chinese) were included in this analysis. Among the 78 English articles, the annual publication volume exhibited fluctuations, peaking in 2024. Chinese institutions and authors led in research output. Diabetes, Metabolic Syndrome, and Obesity had the highest publication volume (7, 8.97%). Keyword identified five clusters: 1) resting-state functional magnetic resonance imaging, 2) metabolic disorders, 3) clinical assessment tools, 4) molecular mechanisms, and 5) emerging fields such as the gut microbiome.Missing dataOnly 22.7% (n = 20) of the studies quantified the missing data, with an average of 9.1%. Among studies with missing data (n = 23), 52.2% (n = 12) provided reasons for missing data, primarily citing poor quality of data collection (41.7%) and loss to follow-up (41.7%). Complete case analysis was the predominant method for addressing missing data (93.3%). No study articulated the hypothesized mechanisms underlying the missing data, and only 4.4% (n = 1) performed a sensitivity analysis.ConclusionIn the domain of T2DM-MCI, research outcomes post-COVID-19 pandemic indicate a rebound, with China maintaining a leading position in scientific research output. However, the reporting of missing data remains ambiguous, and the methods employed to handle such data are insufficient, which may potentially introduce bias.Systematic Review Registrationhttps://doi.org/10.17605/OSF.IO/EZDXM.