Table 7.

Recommendations for the Design and Analysis of Randomized Controlled Trials Involving Longitudinal Data With Irregular Assessment

Recommendation	Reason	Reference
For studies using naturalistic follow-up, include some scheduled assessments for everyone.	Improves robustness of mixed models to informative assessments	30
Measure the outcome variable at baseline.	Can study change from baseline (eliminates random intercepts)	54
Document reasons for irregularity (recommended time to next visit and whether visit is scheduled or as needed).	Reasons for irregularity can help specify a DAG.
Quantify the extent of irregularity (and, in studies with protocolized visit times, include extent of irregularity in DMC reports).	Greater irregularity is associated with greater risk of bias.	33
Specify a DAG, and choose an analytic approach that is appropriate given assumed DAG.	Appropriateness of any given analytic model depends on assumptions made about interrelationships among outcomes and assessment time.	1, 6, 16, 23
Conduct sensitivity analysis.	Cannot ascertain whether the proposed DAG and model are correct; results may be sensitive to mis-specification	1, 30, 50

Abbreviations: DAG, directed acyclic graph; DMC, Data Monitoring Committee.