Table 2.

Study design considerations for development of a NDE

Properties	Questions	Analysis and considerations
Sample size How many patients are needed to answer key questions?	How many patients are needed to detect a treatment effect in an RA population for sleep improvement? Can we detect robust relationships between the reference standard and the wearable sensor parameters?	Power calculation to design study able to detect reference standard (PSG) effect Effect size estimates from the RA and the actigraphy literature can be used here Power calculation to design study able to detect correlation between 2 measures
Test-retest reliability Are parameters measuring values reliably?	Is the ICC for all parameters >0.5? (favored for clinical parameters)	Compute the CV and ICC of each parameter, adjusting for relevant clinical and demographic information Perform at baseline and post treatment
Construct validity: convergent	Variables expected to have positive or negative relationships a priori do in fact follow those expected trends	Correlate parameters with each other (e.g., Spearman correlation)
Construct validity: divergent	Is there a weak relationship between digital sensor and sleep diary scores against other measurements or endpoints that are hypothesized not to be associated with sleep?	Perform K means clustering to determine how many “groups” of variables there are
Construct validity: known groups	Do prespecified hypotheses about variable relationships follow expected trends?
Ability to detect change over the course of treatment	Do changes from baseline values in the treatment arm significantly differ from those in the placebo arm? Are individuals posttreatment significantly different from their pretreatment state?	Standard clinical change from baseline analyses
Establishing meaningful change (e.g., responder/nonresponder)	Are there appropriate criteria to determine responders versus nonresponders?	Determine appropriate cutoffs for drug responders and nonresponders, and test hypotheses

CV, coefficient of variation; ICC, intraclass correlation; RA, rheumatoid arthritis; PSG, polysomnography.