| Title |
The definition of treatment assignment in observational emulations of target trials – an empirical examination in the Swedish Primary Care Cardiovascular Database |
| Authors |
Humphreys, Anna B.C ; Matthews, Anthony A ; Young, Jessica C ; Berglund, Anita ; Lindahl, Bertil ; Wettermark, Björn ; Dahabreh, Issa J ; Kahan, Thomas ; Hernán, Miguel A |
| DOI |
10.1016/j.annepidem.2025.06.003 |
| Full Text |
|
| Is Part of |
Annals of epidemiology.. New York : Elsevier Inc.. 2025, vol. 108, p. 56-62.. ISSN 1047-2797. eISSN 1873-2585 |
| Keywords [eng] |
dispensation ; intention-to-treat effect ; observational data ; prescription ; target trial emulation ; treatment assignment |
| Abstract [eng] |
Background: In randomized trials, the intention-to-treat effect is the effect of assignment to treatment strategies. The concept of assignment may not be clearly defined when using observational data to emulate a target trial. Aims: We aimed to assess the practical implications of using data on prescription versus dispensation as analogues of treatment assignment in observational analyses. Methods: We used the primary care-derived Swedish Primary Care Cardiovascular Database of individuals with newly diagnosed hypertension between 2006 and 2014 and linked registers. We compared the effect of two antihypertensive drug classes on the five-year risk of cancer and ischemic heart disease. Treatment assignment was first mapped using prescription data, and then dispensation data. With unique confounding structures, we sequentially adjusted for different classes of risk factor due to uncertainty over the choice of relevant confounders for prescription vs. dispensation. Results: 7770 individuals were eligible when assignment was defined using prescription compared with 5964 when defined using dispensation. For both cancer and ischemic heart disease outcomes, both higher and lower relative risks of the outcome were consistent with our data. Effect estimates did not vary with the choice of prescription or dispensation data as analogues of assignment, nor with sequential adjustment for class of risk factor. Conclusion: The mapping of prescription or dispensation data to treatment assignment influences the size and characteristics of the study population and the structure of confounding. However, we found no clear numerical differences in effect estimates in this study. Further investigation is required in other settings. |
| Published |
New York : Elsevier Inc |
| Type |
Journal article |
| Language |
English |
| Publication date |
2025 |
| CC license |
|
