Dynamic Treatment Regimens

Structure

A DTR consists of four key ingredients. The first ingredient is a sequence of critical decision points in the medical care process. These decision points may represent time in the form of patient visits to the clinic (first visit, second visit, and so on); or if critical decisions are to be made on a monthly basis, the decision points may represent calendar time in months since disease diagnosis. More generally, though, the sequence of critical decision points is not required to be aligned with a pre-specified set of discrete time points. For example, critical decision points may, instead, be defined by patient events, such as the point at which a patient fails to respond to prior treatment.

The second ingredient is a set of one or more treatment options at each critical decision point. Possible treatment options may be switch medication, augment medication, or continue medication; or there may be more complex options, such as any of the three-way combinations of treatment type (medication, physical therapy), treatment intensity (high, medium, low), and treatment delivery (specialty clinic, general clinic). The set of potential treatment options may differ at different decision points. For example, initially, the emphasis may be on treatment suitable for an acute episode of the illness, whereas subsequent decisions may involve options for intensifying or augmenting treatment for nonresponding patients or transitioning to lower-intensity treatments or monitoring for responding patients.

The third ingredient is a set of one or more tailoring variables at each critical decision point. The tailoring variables form the set of key measures that will determine subsequent treatment. For example, tailoring variables may include patient severity, number and type of comorbidities, side effects resulting from prior treatment, treatment preference, adherence to prior treatment, and, perhaps most important, response to prior treatment. Tailoring variables can also be summary measures over the full course of prior treatment; for example, subsequent treatment could depend on the rate of improvement in symptoms during prior treatment or the pattern of nonadherence to prior treatment. The set of tailoring variables may differ at different time points; for instance, history of comorbidities or genetic background may be used to choose from the options for initial treatment, while the choice of subsequent treatment might be based on response to the present treatment and the type of present treatment.

...