Dear Colleagues:

Does anyone have any favorite articles that focus on the methodology of "change score from baseline" as outcome with baseline value as one covariate, along with demographics and others?

------------------------------
Brandy Sinco, BS, MA, MS
Statistician Senior
Michigan Medicine
------------------------------

If your interest is in issues with interpreting difference scores, I would search out these really old papers by Fred Lord. This topic has been around for a few decades - as I remember, the first paper concerns Lord's Paradox. I don't have copies, sorry. 

Lord FM. A paradox in the interpretation of group comparisons. Psychological
Bulletin 1967; 68: 304-305.
 Lord FM. Statistical adjustments when comparing pre-existing groups.
Psychological Bulletin 1969; 72: 336-337.

------------------------------
Chauncey Dayton
------------------------------

I regard a good wide-ranging text to be:  Garrett Fitzmaurice, Nan M. Laird, James H. Ware. (2011). Applied Longitudinal Analysis (2^nd ed.). Wiley. (ISBN: 978-0-470-38027-7). 

The authors cover this topic in chapter 5, where they discuss four methods of handling baseline.  One of these methods uses "change score from baseline" as an outcome.  Another, uses the baseline value as a covariate.  The authors caution (in Section 5.7) that this latter method "should not be used … in observational studies where there is no a priori reason to assume that the groups have the same mean response at baseline."

 

------------------------------
David Thompson
Emeritus Associate Professor
University of Oklahoma Health Sciences Center
------------------------------

Yes, any article about the regression fallacy. The change = after – before  and baseline = before are correlated even if the before and after are independent. One account is in my article:


------------------------------
Stephen Stigler
The University of Chicago
------------------------------

If randomization was used to create the groups of analytic units for comparison of outcome then consider the following:

Guanghan F. Liu∗,†, Kaifeng Lu, Robin Mogg, Madhuja Mallick and Devan V. Mehrotra: Should baseline be a covariate or dependent variable in analyses of
change from baseline in clinical trials? Statist. Med. 2009; 28:2509–2530

------------------------------
Brent Blumenstein
------------------------------

Like a) Frison&Pocock 1992 StatMed for RCT or b) Tennant et al. 2021 IntJEpi for causality in observational data? As c) SEM by Kievit et al. 2018 DevelopmentCognitNeurosci? Or d) why marginal odds ratios are different from conditional odds ratios (Daniel et al. 2021 BiomJ)? a) and c) assume linear model. d) contrasts this with GLM.

------------------------------
Reinhard Vonthein
Universitaet zu Luebeck
------------------------------

Senn, S. (2006). Change from baseline and analysis of covariance revisited. Stat Med, 25(24):4334–44.


------------------------------
Ralph O'Brien
Professor of Biostatistics (officially retired; still keenly active)
Case Western Reserve University
http://rfuncs.weebly.com/about-ralph-obrien.html
------------------------------