gms | German Medical Science

67. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e. V. (GMDS), 13. Jahreskongress der Technologie- und Methodenplattform für die vernetzte medizinische Forschung e. V. (TMF)

21.08. - 25.08.2022, online

The handling of missing values in diagnostic studies

Meeting Abstract

Suche in Medline nach

  • Katharina Stahlmann - Department of Medical Biometry, University Medical Center Hamburg-Eppendorf, Hamburg, Germany
  • Antonia Zapf - Department of Medical Biometry, University Medical Center Hamburg-Eppendorf, Hamburg, Germany

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie. 67. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e. V. (GMDS), 13. Jahreskongress der Technologie- und Methodenplattform für die vernetzte medizinische Forschung e.V. (TMF). sine loco [digital], 21.-25.08.2022. Düsseldorf: German Medical Science GMS Publishing House; 2022. DocAbstr. 112

doi: 10.3205/22gmds087, urn:nbn:de:0183-22gmds0878

Veröffentlicht: 19. August 2022

© 2022 Stahlmann et al.
Dieser Artikel ist ein Open-Access-Artikel und steht unter den Lizenzbedingungen der Creative Commons Attribution 4.0 License (Namensnennung). Lizenz-Angaben siehe http://creativecommons.org/licenses/by/4.0/.


Gliederung

Text

Introduction: Missing values always pose a threat to study validity and can introduce bias [1], [2], [3]. In diagnostic accuracy studies, this can have severe consequences by leading to mis-diagnosing a large number of patients and assigning them to the wrong treatment [4]. Although the STARD guideline clearly recommends to report the handling of missing values and inconclusive results [2], little research has been conducted on appropriate methods – especially regarding missing values in the index test. Methods for dealing with missing data in therapeutic studies cannot simply be applied to diagnostic studies, as these have distinct characteristics. For instance, they have the co-primary endpoints sensitivity and specificity and a different design, in which all patients undergo the reference, index and – if applicable – a comparator test [5]. Consequently, this project aimed at providing an overview of existing strategies to handle missing values in diagnostic accuracy studies, comparing them, identifying research gaps and proposing necessary steps for further research.

Methods: As part of a position paper, a structured literature review was conducted searching the databases MEDLINE, Cochrane Library and Web of Science to identify papers addressing the handling of missing values or inconclusive, indeterminate, uninterpretable or intermediate results in either the reference, gold standard or index test within a diagnostic study. The focus was on methodological articles but clinical trials were also included if they discussed the performance of a missing data method. The proposed methods in these papers were compiled, compared and evaluated regarding their performance to handle missing values in diagnostic studies.

Results: The majority of the finally 94 identified studies presented strategies to handle missing values in the reference / gold standard while very few studies discussed missing values or inconclusive results in the index test. The discussed methods comprised complementing with external data, using multiple tests, multiple imputation (MI), Bayesian and likelihood based approaches, latent class analysis and inverse probability weighting. Overall, MI was the most recommended strategy as it outperformed simple methods, such as single imputation and complete case analysis. Nearly all studies explored the performance of the proposed methods only under MCAR or MAR.

Discussion: A range of different methods to handle missing values in diagnostic studies could be identified. However, most studies investigated missing values in the reference / gold standard and under MCAR and MAR. Few examined the performance of methods under MNAR or missing data in the index test. Moreover, the studies compared only some selected methods instead of all possible approaches systematically.

Conclusion: Future research should systematically compare and evaluate all possible methods for handling missing values in the index test under different scenarios, including MNAR, to be able to make well-founded recommendations. This review contributes to the current literature by providing an overview of existing strategies on handling missing data in diagnostic studies and by identifying research gaps and the necessary next steps regarding this topic.

The authors declare that they have no competing interests.

The authors declare that an ethics committee vote is not required.


References

1.
EMA. Guideline on Missing Data in Confirmatory Clinical Trials. London, UK: European Medicines Agency; 2011.
2.
Cohen JF, Korevaar DA, Altman DG, Bruns DE, Gatsonis CA, Hooft L, et al. STARD 2015 guidelines for reporting diagnostic accuracy studies: explanation and elaboration. BMJ Open. 2016;6(11):e012799. DOI: 10.1136/bmjopen-2016-012799 Externer Link
3.
Schuetz GM, Schlattmann P, Dewey M. Use of 3x2 tables with an intention to diagnose approach to assess clinical performance of diagnostic tests: meta-analytical evaluation of coronary CT angiography studies. BMJ. 2012;345:e6717. DOI: 10.1136/bmj.e6717 Externer Link
4.
Philbrick JT, Horwitz RI, Feinstein AR, Langou RA, Chandler JP. The limited spectrum of patients studied in exercise test research. Analyzing the tip of the iceberg. JAMA. 1982;248(19):2467-70.
5.
EMA. Guideline on Clinical Evaluation of Diagnostic Agents. London: European Medical Agency; 2010.