pg 13 makes a point of considering follow up time in observational studies and makes the comment:
This should immediately raise a red flag about the study above, because the unexposed group is defined as never receiving pioglitazone during the follow-up period.
are there study designs for observational data and treatment effects that better deal with this issue; such as counting all person time when not on drug and when on drug during a specified period of time for each person?
in your example, you excluded the person time that occurred before the patient started the drug during that 8 year period
could you expound on study designs that are better than the example you provided for situations in which you cannot do IV, falsification tests or propensity analyses?