Catharine B. Stack, PhD, MS; A. Russell Localio, PhD; Michael E. Griswold, PhD; Steven N. Goodman, MD, PhD, MHS; Cynthia D. Mulrow, MD, MSc
Acknowledgment: The authors thank Vasilakou and colleagues for providing the analytic data set from Figure 2 of their article.
Potential Conflicts of Interest: Disclosures can be viewed at www.acponline.org/authors/icmje/ConflictOfInterestForms.do?msNum=M13-1657.
Requests for Single Reprints: Catharine B. Stack, PhD, MS, American College of Physicians, 190 N. Independence Mall West, Philadelphia, PA 19106; e-mail, cstack@acponline.org.
Current Author Addresses: Dr. Stack: American College of Physicians, 190 N. Independence Mall West, Philadelphia, PA 19106.
Dr. Localio: Center for Clinical Epidemiology and Biostatistics, University of Pennsylvania, 635 Blockley Hall, 423 Guardian Drive, Philadelphia, PA 19104-6021.
Dr. Griswold: Center of Biostatistics, University of Mississippi Medical Center, 2500 North State Street, Jackson, MS 39216.
Dr. Goodman: Stanford University School of Medicine, 259 Campus Drive, T265 Redwood Building/HRP, Stanford, CA 94305.
Dr. Mulrow: University of Texas Health Science Center at San Antonio, 7703 Floyd Curl Drive, San Antonio, TX 78229.
Author Contributions: Conception and design: C.B. Stack, A.R. Localio, M.E. Griswold.
Analysis and interpretation of the data: C.B. Stack, A.R. Localio, M.E. Griswold, S.N. Goodman, C.D. Mulrow.
Drafting of the article: C.B. Stack, A.R. Localio, M.E. Griswold, S.N. Goodman, C.D. Mulrow.
Critical revision of the article for important intellectual content: C.B. Stack, A.R. Localio, M.E. Griswold, S.N. Goodman, C.D. Mulrow.
Final approval of the article: C.B. Stack, A.R. Localio, M.E. Griswold, S.N. Goodman, C.D. Mulrow.
Statistical expertise: C.B. Stack, A.R. Localio, M.E. Griswold, S.N. Goodman.
Collection and assembly of data: C.B. Stack.
Confidence in evidence summarized in meta-analyses depends on the strength of the underlying studies. This inherent limitation of syntheses appears in the case of a meta-analysis of sodium–glucose cotransporter 2 inhibitors for the treatment of type 2 diabetes because many of the pertinent randomized trials did not handle patient dropout and “rescue” medication properly. Repudiated statistical methods, such as last observation carried forward, and unsophisticated methods for handling postrescue data produce unreliable summary estimates. Future reports of randomized studies and meta-analyses of those studies must focus on posing precise questions about the treatment effect of interest and then implement appropriate statistical methods to account for missing data, patient dropout, and use of rescue medication.
Stack CB, Localio AR, Griswold ME, Goodman SN, Mulrow CD. Handling of Rescue and Missing Data Affects Synthesis and Interpretation of Evidence: The Sodium–Glucose Cotransporter 2 Inhibitor Example. Ann Intern Med. ;159:285–288. doi: 10.7326/0003-4819-159-4-201308200-00009
Download citation file:
© 2019
Published: Ann Intern Med. 2013;159(4):285-288.
DOI: 10.7326/0003-4819-159-4-201308200-00009
Research and Reporting Methods.
Results provided by: